Fix bug when fetching normalized interactions as dense matrices

[robomics]

This PR addresses a bug detected by the new fuzzer suite.
In brief, fetching normalized pixels as dense matrices could sometime results in incorrect results, where some entries that were supposed to be inf (due to e.g. normalization with divisive weights of value 0) were instead set to nan.

[codecov]

Codecov Report

Attention: Patch coverage is 73.56322% with 23 lines in your changes missing coverage. Please review.

Project coverage is 74.54%. Comparing base (365f138) to head (a399491).

Files with missing lines	Patch %	Lines
...e/hictk/transformers/impl/to_dense_matrix_impl.hpp	64.70%	9 Missing and 3 partials ⚠️
...hictk/hic/include/hictk/hic/impl/hic_file_impl.hpp	83.33%	4 Missing and 1 partial ⚠️
src/libhictk/file/include/hictk/impl/file_impl.hpp	72.72%	2 Missing and 1 partial ⚠️
.../hictk/transformers/impl/to_sparse_matrix_impl.hpp	50.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #252      +/-   ##
==========================================
+ Coverage   74.52%   74.54%   +0.01%     
==========================================
  Files         144      144              
  Lines       12434    12451      +17     
  Branches     1445     1439       -6     
==========================================
+ Hits         9266     9281      +15     
- Misses       1989     1993       +4     
+ Partials     1179     1177       -2     

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