More intelligently search for the right post-processed files

[andrew-c-ross]

To support newer simulations, this changes the behavior of open_var to look for the right post-processed files, without any hard coding of the dates or chunk lengths.

The general assumptions are

1. A path following the standard frepp naming, for example {pp_root}/pp/ocean_monthly/ts/monthly/, exists (error if not)
2. Within that path, one or more directories with names like *yr exist (error if not)
3. Of the directories above, whichever has the largest number in the name and contains one or more post-processed files of the right variable is the one to use.

We generally post-process two different chunk sizes, one of 5 years and the other the length of the simulation. This should work for that. One way to get a suboptimal result would be if the simulation length was longer than the longest chunk; for example, a 25 year simulation with 5 year and 20 year chunks. This is an unlikely user error though.

[uwagura]

ran it successfully with my own 27 year run, looks good to me!

[yichengt900]

Thanks again, @andrew-c-ross, for your contribution! Just being a little picky about tiny details, but everything looks perfect to me!