Fix ValueError when using string norm with vmin/vmax

[Ai-chan-0411]

Summary

When calling plotting methods like pcolormesh with a string normalizer (e.g. norm="log") and explicit vmin/vmax limits, the current code raises ValueError: If 'norm' is given, 'vmin' and 'vmax' must not be set. even though this combination is perfectly valid — the string norm needs to be constructed with those limits.

I ran into this while trying to plot log-scaled data with a custom lower bound (vmin=1e-2). The guard in _parse_cmap was checking norm is not None when it should only reject the case where norm is an already-instantiated Normalize object (which carries its own vmin/vmax internally). When norm is a string or tuple, it gets passed to constructor.Norm() later, which accepts vmin/vmax as constructor arguments.

The fix narrows the check to isinstance(norm, mcolors.Normalize) so that string-based norms can coexist with explicit limits. I also added a regression test that confirms pcolormesh(data, norm="log", vmin=1e-2) works without error and the resulting normalizer has the expected vmin.

Fixes #689

[cvanelteren]

#662

[Ai-chan-0411]

Thanks for reviewing and closing this! I appreciate you pointing to PR #662 as the preferred approach for handling this ValueError issue. I'll keep that pattern in mind for future contributions.