Fix common normalization for density histograms

[yushuosun]

Hist._eval currently asks NumPy to normalize each semantic group independently when stat="density". By the time seaborn applies its common_norm grouping, the original counts have already been lost.

This keeps raw histogram counts in _eval and moves density normalization into _normalize, alongside the other normalized statistics. As a result:

• the combined area is 1 with the default common_norm=True;
• each group has area 1 with common_norm=False;
• list-valued common_norm normalizes within each requested subset.

The full _stats test suite passes (74 passed, 2 skipped), including three regression tests added here.

This revives the approach from #3911, which was withdrawn by its author while the issue remained open. Thanks @Mr-Neutr0n for the original diagnosis and test cases.

Closes #3633.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.