Fix GR text-state corruption causing persistent layout assertion failures

[AJ0070]

Fixes a GR backend state-corruption bug behind issue #4816 where plotting with non-finite text coordinates/extents could poison subsequent plots in the same session and trigger layout assertions (total_plotarea_vertical > 0mm / total_plotarea_horizontal > 0mm).

Root cause

When GR was asked to draw or measure text at invalid coordinates (for example, annotate! with NaN positions, or labels/ticks after invalid log-scale inputs), GR could return non-finite text extents (inqtext / inqtextext). Those values propagated into subplot padding and layout calculations, causing assertion failures. Because GR state persisted, later unrelated plots also failed until restart.

Fix

In GRExt.jl:

• gr_text(...): skip text draw when (x, y) is non-finite.
• gr_inqtext(...): sanitize non-finite returned text bounds and fall back to finite zero extents.

This prevents non-finite values from entering layout padding math while preserving normal behavior for valid inputs.

Testing

Added regression test in test_misc.jl covering:

• Non-finite annotation reproducer (including repeated call and history dependence)
• Invalid negative log-scale case followed by a valid labeled log-scale plot

Strengthened test path to renderer-level checks (show(..., MIME("image/png"), ...)) so GR drawing and measuring is exercised directly.

Additional GR stress verification in a clean environment (repeated bad → good render cycles, including twinx) completed without reintroducing assertion failures or session poisoning.

Attribution

• I am listed in the appropriate version of .zenodo.json for PRs against v2 or PRs against master (see How to get acknowledged #3503)

Things to consider

• Does it work on log scales?
• Does it work in layouts?
• Does it work in recipes?
• Does it work with multiple series in one call?
• PR includes or updates tests?
• PR includes or updates documentation?

[AJ0070]

The code block mentioned in the issue description works fine locally but there is failure on CI. I am trying to figure it out. It would be great if someone could help out.