lksearch should return either a file in the cache or an HDUList in memory

[christinahedges]

At the moment we have lksearch only returning files stored to disk, which was our first milestone. We now need a way to have lksearch return either a path to a file, or an astropy.fits.HDUlist in memory. Here are some scenarios;

• User wants a TPF/LCF from Kepler/K2/TESS, wants and wants to cache that data for future use to prevent more queries. File should should be cached on disk. User could request either path to file, or astropy.fits.HDUList object in memory.
• User wants a cut out TPF from the cubes, wants to cache that for future use to prevent more queries. File should be cached on disk. User could request either path to file, or astropy.fits.HDUList object in memory.
• User wants a TPF/LCF/cut out TPF but does not need to cache it for the future (either won't need it, or are working on TiKE/AWS). File should not be cached on disk. User should only get an astropy.fits.HDUList object in memory.

We need the following

• lksearch should have an option in the configuration on whether to cache files on disk. If no, lksearch would download data everytime. (This option needs )
• The download functions in lksearch should have an option for what to return, either paths as a pd.DataFrame or an astropy.fits.HDUList.
• In the event that any part of the download failed and the file is corrupted, download should raise an exception.
• Using the configuration system, lksearch should be able to configure whether the default output of search is paths as a pd.DataFrame or an astropy.fits.HDUList. Sane default is probably astropy.fits.HDUList.
• Using the configuration system, lksearch should be able to configure whether the default option is to cache files or not.

@tylerapritchard @Nschanche what's your take on this? Some of this might already exist inside lksearch.