fix(text): raise a clear error instead of IndexError on glyph/char mismatch

[SupernovaIa]

Summary

Text._gen_chars() assumes disable_ligatures=True guarantees exactly one rendered glyph per non-space character. That assumption breaks for fonts that implement some of their ligatures through OpenType features other than liga/dlig/clig/hlig (the ones ManimPango's disable_liga disables) — for example, Fira Code implements its programming ligatures (<=, ->, etc.) exclusively through calt (contextual alternates), which isn't covered. I confirmed this by inspecting Fira Code's own GSUB table with fontTools: it defines calt but none of liga/dlig/clig/hlig.

When that happens, Pango still merges e.g. < and = into a single glyph, and _gen_chars() walks off the end of self.submobjects, raising an opaque IndexError: list index out of range with no indication of what went wrong. This is reproducible today via, for instance:

Code(
    code_string="print(1<=2)",
    language="python",
    paragraph_config={"font": "Fira Code"},
)

This PR does not fix the underlying issue — the actual root cause is in a separate repository, ManimCommunity/ManimPango (its disable_liga feature string would need calt=0 added). That's outside the scope of this PR (would require setting up the Cython/C build for a separate package).

What this PR does do: turn the opaque crash into a clear, actionable error, so that anyone hitting this (there have been a few confused reports in #3237 over time) immediately understands why, instead of debugging a bare IndexError. This is a pure improvement to the failure path — no behavior changes for the cases that already work.

Implementation note

The check is a bounds guard inside the glyph-consuming loop, raised only when there are genuinely fewer glyphs than non-space characters. An earlier version of this PR used a pre-loop count that compared non-space characters against the total number of submobjects; that was wrong, because under disable_ligatures=True a space is rendered as its own glyph, so the two counts legitimately differ for any text containing spaces. It produced a false positive on ordinary spaced text — including the DisableLigature docs example (Text("fl ligature", disable_ligatures=True)), which broke the docs build. The in-loop bounds check fires only on the real glyph shortfall.

Refs #3237

Test plan

• test_gen_chars_raises_clear_error_on_glyph_mismatch — simulates the glyph shortfall directly (no dependency on any specific font being installed) and asserts the new ValueError is raised
• test_gen_chars_handles_spaces_with_disable_ligatures — renders Text("fl ligature", disable_ligatures=True) (the case that broke the docs build) and asserts it does not raise
• uv run pytest tests/module/mobject/text/test_text_mobject.py tests/test_code_mobject.py — all pass
• Manually reproduced the original crash with Fira Code installed locally, confirmed the new error message is accurate and points at the real cause
• Rebuilt the DisableLigature docs example locally, confirmed it renders without error
• uv run pre-commit run --files manim/mobject/text/text_mobject.py tests/module/mobject/text/test_text_mobject.py — ruff, mypy, codespell all pass