Convert counts to integer to avoid floating point weirdness.

[schlafly]

This PR clips the number of photons entering a pixel over an exposure to 2 * 10**9 to avoid floating point / integer math issues. This seems like a lot of photons, but a 0th mag star should generate O(10^11) photons per second per meter squared, so it's not that many. Roman saturates at O(10^5) photons and we really only need to care about handling enough to saturate in a single read; this lets us handle exposure times up to ~2*10^4 reads which seems like more than enough. The non-linearity correction will have issues long before this and would be the next thing to look at if we wanted to do a better job with very bright stars.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.73%. Comparing base (725b7cb) to head (103d9c2).
Report is 14 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #119      +/-   ##
==========================================
- Coverage   92.44%   90.73%   -1.71%     
==========================================
  Files          16       16              
  Lines        1509     1630     +121     
==========================================
+ Hits         1395     1479      +84     
- Misses        114      151      +37     

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