Fix typo and document normalize values in confusion_matrix README

[RudrenduPaul]

Fix typo and document normalize values in confusion_matrix README

Closes #652

What this fixes

In metrics/confusion_matrix/README.md:

1. Typo in Output Values: "or the number of examples input" is missing
   a word and should read "or the number of examples in the input". The
   same sentence also referenced normalize being set to True, but
   normalize is a string parameter (passed straight through to
   sklearn.metrics.confusion_matrix) and its default is None, not a
   boolean True. Corrected to "if normalize is set to None (default)".

2. Missing valid values for normalize: The Inputs section described
   what normalize does but never listed the permitted values. Added the
   permitted values 'true', 'pred', 'all', or None (default),
   matching the underlying sklearn.metrics.confusion_matrix implementation
   used by this metric (see metrics/confusion_matrix/confusion_matrix.py,
   which passes normalize straight through to sklearn).

Verification

• Confirmed against current main that both issues were still present
  before this fix.
• Change is scoped to the README only; no code or behavior changes.

Note: this documentation fix was prepared with AI assistance (Claude Code)
and reviewed before submission.