Fix performance issues in float parsing.

[Alexhuszagh]

This fixes a few performance issues due to a lack of inlining on the Eisel-Lemire algorithm.

[Alexhuszagh]

According to my benchmarks, the ones in fast-float2 against the latest commit are:

Canada

=====================================================================================
|                         canada.txt (111126, 1.93 MB, f64)                         |
|===================================================================================|
|                                                                                   |
| ns/float                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2           10.37    10.45    10.70    10.98    11.36    13.17    26.49 |
| fast-float             9.96    10.18    10.31    10.46    11.02    12.34    16.35 |
| lexical               10.40    10.49    10.52    10.65    11.13    11.88    14.43 |
| from_str              11.56    11.72    11.84    11.96    12.48    13.12    25.68 |
|                                                                                   |
| Mfloat/s                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2           37.75    76.05    88.03    91.09    93.46    95.73    96.46 |
| fast-float            61.16    81.04    90.72    95.57    97.02    98.23   100.40 |
| lexical               69.32    84.26    89.89    93.90    95.03    95.37    96.12 |
| from_str              38.94    76.22    80.11    83.61    84.48    85.31    86.48 |
|                                                                                   |
| MB/s                    min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2          656.91  1323.30  1531.80  1585.04  1626.36  1665.87  1678.60 |
| fast-float          1064.19  1410.26  1578.57  1663.01  1688.27  1709.31  1747.15 |
| lexical             1206.25  1466.18  1564.27  1634.06  1653.62  1659.58  1672.64 |
| from_str             677.53  1326.30  1394.09  1454.93  1470.08  1484.53  1504.86 |
|                                                                                   |
=====================================================================================

Mesh

=====================================================================================
|                          mesh.txt (73019, 0.54 MB, f64)                           |
|===================================================================================|
|                                                                                   |
| ns/float                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2            5.02     5.13     5.22     5.45     5.91    10.35    18.15 |
| fast-float             4.83     4.94     5.01     5.17     5.31     6.20    10.26 |
| lexical                4.83     4.90     4.97     5.04     5.39     8.44    10.61 |
| from_str               5.78     5.86     5.93     6.00     6.10     6.63    10.04 |
|                                                                                   |
| Mfloat/s                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2           55.10    97.33   169.50   183.56   191.70   194.98   199.02 |
| fast-float            97.48   163.54   188.44   193.27   199.56   202.61   206.85 |
| lexical               94.25   118.83   185.70   198.58   201.38   204.02   206.85 |
| from_str              99.62   151.05   163.90   166.71   168.56   170.84   173.15 |
|                                                                                   |
| MB/s                    min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2          404.44   714.49  1244.22  1347.43  1407.22  1431.27  1460.91 |
| fast-float           715.54  1200.47  1383.25  1418.76  1464.91  1487.26  1518.44 |
| lexical              691.89   872.27  1363.20  1457.73  1478.24  1497.65  1518.44 |
| from_str             731.25  1108.83  1203.16  1223.76  1237.32  1254.12  1271.07 |
|                                                                                   |
=====================================================================================

Random Uniform

=====================================================================================
|                           uniform (50000, 0.87 MB, f64)                           |
|===================================================================================|
|                                                                                   |
| ns/float                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2            9.61     9.78     9.87    10.44    11.03    14.03    25.82 |
| fast-float             9.30     9.38     9.52    10.48    10.66    12.17    18.72 |
| lexical                9.44     9.54    10.01    10.78    11.03    13.55    17.52 |
| from_str              10.67    10.78    10.89    11.05    11.97    14.30    21.30 |
|                                                                                   |
| Mfloat/s                min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2           38.73    71.34    90.68    95.82   101.32   102.25   104.01 |
| fast-float            53.42    82.22    93.86    95.42   105.04   106.59   107.55 |
| lexical               57.07    73.78    90.65    92.82   100.18   104.82   105.93 |
| from_str              46.95    70.01    83.56    90.51    91.86    92.75    93.69 |
|                                                                                   |
| MB/s                    min       5%      25%   median      75%      95%      max |
|-----------------------------------------------------------------------------------|
| fast-float2          674.92  1243.06  1580.09  1669.72  1765.47  1781.72  1812.48 |
| fast-float           930.83  1432.76  1635.56  1662.71  1830.38  1857.30  1874.08 |
| lexical              994.48  1285.61  1579.52  1617.34  1745.66  1826.54  1845.89 |
| from_str             818.09  1219.91  1455.98  1577.23  1600.70  1616.14  1632.49 |
|                                                                                   |
=====================================================================================

Other benches in lexical show similar results (one uses criterion, one is a straight executable), where the performance is practically identical for fast-cast datasets, but otherwise lexical outperforms for near-halfway cases. The performance for the Earth dataset is particularly a lot faster than fast-float2, which was unexpected since it's very simple floats like to be covered by the fast case (can be represented entirely by machine floats).