#StartFloat syntax

[cbmarini]

This PR updates the #StartFloat syntax as discussed in #676. The new syntax is

#StartFloat <precision>, MZV = <weight>

where the precision can either be in bits (b) or digits (d). For example:

#StartFloat 50b, MZV = 2
#StartFloat 20d, MZV = 3

Notes:

• To properly test whether the correct number of digits are printed, PR Fix float printing #676 should be merged first.
• In this regard, it is also worth reviewing whether the two separate calls to SetFloatPrecision from SetupMZVTables and SetupMPFTables are both needed (both increase the precision by the MZV weight). Likely, a single call in e.g. DoStartFloat should be sufficient.

[jodavies]

This looks like a good change, but remember to also update the manual.

I would perhaps also merge the commits, since the CI will not pass on the first commit alone, and perhaps makes bisection a bit less smooth.

* In this regard, it is also worth reviewing whether the two separate calls to `SetFloatPrecision` from `SetupMZVTables` and `SetupMPFTables` are both needed (both increase the precision by the MZV weight). Likely, a single call in e.g. `DoStartFloat` should be sufficient.

I think this is also a good idea, but maybe is better as part of #676 which is already changing how the allocation and precision is handled.

[coveralls]

coverage: 50.595% (-0.008%) from 50.603%
when pulling 1031e1d on cbmarini:StartFloat
into 21f29a8 on form-dev:master.

[cbmarini]

I updated the manual and merged the commits. Do we also want to add the new preprocessor commands and statements that come with the floating point system added earlier in the respective chapters?

[jodavies]

Yes I think so, at least in the style of "see the Floating point chapter", similar to optimize: https://www.nikhef.nl/~form/maindir/documentation/reference/html/manual.html#x1-550003.38

[cbmarini]

Allright, I will update the manual. At the moment the manual says that the first precision argument of #StartFloat is mandatory. However, at the moment it is actually optional and if you don't specify a precision, FORM will fall back onto DEFAULTPRECISION which is set to 1000. Do we want this to be optional with indeed 1000 bits of precision as the default?

[cbmarini]

Furthermore, with the new syntax, one could also relax the order of the two arguments in #StartFloat.