Use scipy.sparse for histogram storage.

[delucchi-cmu]

Closes #154 using scipy.sparse for representation of healpix histogram.

Using the dense np array on-disk, the intermediate histograms for TIC are 8.5 G. Using sparse array, it's 48 M (around 200X improvement in disk usage). There's no discernible performance penalty.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.52%. Comparing base (7a5c17c) to head (ad989ba).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #294      +/-   ##
==========================================
+ Coverage   99.36%   99.52%   +0.16%     
==========================================
  Files          24       25       +1     
  Lines        1260     1273      +13     
==========================================
+ Hits         1252     1267      +15     
+ Misses          8        6       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jeremykubica]

Looks good. I added a bunch of optional comments and thoughts, but nothing blocking.