Use R_READABS_PATH as a cache

[HughParsonage]

We could use the path R_READABS_PATH:

1. to cache the results of read_abs as a .fst file, to improve performance, and perhaps harness the random-access features of that file.
2. using the modification time, the maximum date in the file, or, say, a file containing the MD5 hashes, to determine whether the local file needs to be updated when reading. (add ability to check if local file exists & is up to date #46 )

The eventual goal is for the user to just type read_abs("6345.0") and for readabs to:

1. if the file exists and is up-to-date---simply read it immediately
2. if the file does not exist or is not up-to-date---download the .xls file, tidy it, and then read the tidy files immediately on subsequent occasions.

without the user having to tradeoff performance against accuracy/recency.

[MattCowgill]

I like the idea @HughParsonage!

The ABS Time Series Directory (the crummy old API that the package is built around) does return a value for "Frequency" and "ProductReleaseDate" (which isn't always the same as the modification date) for each time series. We could use those to figure out whether the file is (likely to be) up to date

[HughParsonage]

Cool. Could I get push access to this repo? Makes it easier to collab rather than messing about with forks. Don't need master branch push permission.

[MattCowgill]

Done

Let me know what you're thinking!