Fix Division-by-Zero in Masked Statistical Functions

[JAi-SATHVIK]

Masked statistical functions (mean, variance, correlation, and covariance) currently crash or return NaN when provided with an "all-false" mask. This is caused by a missing guard against zero counts in denominators specifically when count(mask) is 0 leading to critical division-by-zero errors during runtime.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 68.53%. Comparing base (8116e5c) to head (88f13a0).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1122      +/-   ##
==========================================
- Coverage   68.55%   68.53%   -0.02%     
==========================================
  Files         396      396              
  Lines       12746    12746              
  Branches     1376     1376              
==========================================
- Hits         8738     8736       -2     
- Misses       4008     4010       +2     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jalvesz]

Instead of introducing another internal implicit loop (using count) it might be better to store the denominator in its own local variable and check whether the denominator is equal to zero or not to assign NaN. E.g:

denom = dot_product( centeri_, centeri_)*dot_product( centerj_, centerj_)
if (denom<epsilon(0._${k1}$)) then
     res(j, i) = ieee_value(1._${k1}$, ieee_quiet_nan)
else
     res(j, i) = dot_product( centerj_, centeri_)/sqrt(denom)
end if

same applies elsewhere