Guard against rules that don't advance state.line / state.pos, as markdown-it JS does

[acaprino]

Context

Since 13.0.2 (markdown-it/markdown-it#847), markdown-it JS stops with an error when a rule reports a match without consuming input, instead of looping:

• ParserBlock.tokenize throws block rule didn't increment state.line (parser_block.ts);
• ParserInline.skipToken and ParserInline.tokenize throw inline rule didn't increment state.pos (parser_inline.ts).

markdown-it-py has no equivalent check in ParserBlock.tokenize or in ParserInline.skipToken / tokenize. A plugin rule that returns True without advancing makes parse() loop forever. It usually pushes a token on every pass, so memory grows until the process is killed:

from markdown_it import MarkdownIt

md = MarkdownIt("commonmark")
md.block.ruler.before("paragraph", "stuck", lambda state, start, end, silent: True)
md.parse("text\n")  # never returns; same with an inline rule that returns True

This is not hypothetical. executablebooks/mdit-py-plugins#156 (texmath, input "> $$ a=1\n\n$$\n") and, earlier, executablebooks/mdit-py-plugins#117 (amsmath) were both this bug in a plugin. In each case the only symptom was a hang.

Proposal

Port the JS guards: after a rule returns True, raise if state.line (block) or state.pos (inline) has not advanced. A buggy plugin would then fail fast with a clear message that names the problem, instead of hanging and exhausting memory. Built-in rules always advance, so this only affects plugins that are already broken.

Reproduced on markdown-it-py 4.2.0 and master (6f58654), Python 3.12. I'm happy to open a PR if you'd like one.