pspecdata input calibration and pspec_run updates

[nkern]

General updates in support of H1C_IDR2 pipeline overhaul. This includes:

• allows PSpecData.add() to take a calibration file and apply on-the-fly

• new interleave_times kwarg to pspec_run

• allow for _load_dsets to take multiple files for a single entry in dsets

• pspec_run now only returns ds instead of psc, ds

• ds.validate_datasets bug fix when doing auto-dset power spectra

• optional SWMR mode in PSpecContainer

• minor changes to grouping.bootstrap_run to incorporate PSpecContainer transactional mode

[coveralls]

Coverage increased (+0.02%) to 96.68% when pulling 8eff919 on pspecdata_apply_cal into aca6586 on master.

[nkern]

@plaplant okay I added a print warning and updated a docstring in PSpecContainer in container.py. Let me know if you have any further comments on SWMR usage in PSpecContainer (no need to review the whole PR if you are pressed for time) thanks!

[plaplant]

@nkern I think the way forward here is to run a check each time a new group or dataset would be added to a pspec file, and raise an error if the file is in SWMR mode. I agree that I don't want to hamstring the user too much, but also want to make sure that we're using the library correctly. (Re: the issue you posted yesterday, the library as currently written seems to be able to create datasets without error, but I am skeptical the data are actually being written correctly. Even if you showed it was not writing nonsense for a few datasets, I don't think we should be doing it at all because the library makes no guarantees to correctness, and places all responsibility on the user.)

I think defaulting to swmr=False (as I think you've done here) and having the aforementioned check would be sufficient to make things "safe enough" for the typical user. Thanks for adding the documentation, that's also very helpful for the user to understand the limitations. I can take a look, but probably can't get to it for another ~week or two.

[nkern]

@philbull @plaplant Okay, I've updated PspecContainer such that it will not write new groups or datasets if SWMR is turned on. This necessitated changing some of the testing structure in test_pspecontainer.py, but I think everything is consistent now. @philbull feel free to review the PR now and @plaplant let me know if you have any more concerns on SWMR usage in PSpecContainer (again no need to review the whole PR, Paul). Thanks!

[philbull]

Looks good, just a few minor comments on docstrings and a query about whether to move some utility functions.