Use the shared finite-scalar validator for axis tick values

[Alek99]

Summary

Most numeric builder arguments use XY's canonical validation rule: reject booleans, coerce numeric scalar-likes, reject NaN/infinity, and raise a contextual ValueError. x_axis() and y_axis() bypass that rule for major and minor tick values by calling float() directly.

Audited at 99eda6d.

Current behavior

Both axis factories convert tick_values and minor_tick_values with raw list comprehensions in x_axis() and y_axis().

Consequences include:

• tick_values=[True] is accepted as [1.0];
• NaN and positive/negative infinity are accepted into a public axis spec;
• invalid values leak raw float() conversion messages rather than a stable error naming the axis and tick index.

The repository already has the intended primitive in _validate.finite_scalar.

Proposed direction

Add a small shared sequence validator that applies finite_scalar with indexed labels, and use it for major and minor tick values on both axes (and any equivalent polar-axis input that shares the same public contract).

Acceptance criteria

• Major and minor tick values reject Python/NumPy booleans.
• Major and minor tick values reject NaN and infinities before compilation.
• Conversion failures raise a contextual ValueError naming x_axis/y_axis, the argument, and ideally the index.
• Valid Python numbers and NumPy numeric scalar-likes continue to work.
• Tests cover x/y, major/minor, bool, non-finite, invalid-string, and valid NumPy-scalar inputs.