IndexError: table inside a blockquote at end of input (html_block.py, heading.py)

[WilcoPancras]

Describe the bug

Input that ends on a blockquote marker while a table is open inside that quote raises
IndexError: string index out of range.

Two block rules are affected, both reached as terminator rules from rules_block/table.py:190:

rule	line
rules_block/html_block.py	if state.src[pos] != "<":
rules_block/heading.py	`ch: str

Both compute pos = state.bMarks[startLine] + state.tShift[startLine], which equals len(state.src)
for the empty final line a trailing > produces. In markdown-it (JS) the equivalent is
state.src.charCodeAt(pos), which returns NaN out of range instead of throwing — the port hazard
tracked in #190.

Sibling rules already defend against exactly this: hr.py and blockquote.py wrap the same index in
try: ... except IndexError: return False (added for #185 / #204). html_block.py and heading.py
appear to have been missed.

html_block shadows heading because it runs first, so disabling html moves the traceback rather
than fixing it.

Reproduce the bug

from markdown_it import MarkdownIt

# html_block.py:47
MarkdownIt().enable("table").parse("> | a | b |\n> |---|---|\n>")

# heading.py:22 (same input, html_block bails at its options check first)
MarkdownIt("commonmark", {"html": False}).enable("table").parse("> | a | b |\n> |---|---|\n>")

Both raise:

IndexError: string index out of range

Needs table enabled, so gfm-like and js-default are affected; plain commonmark is not, since
without the table rule the terminator rules are never run on that line.

Variants (all with MarkdownIt().enable("table"))

input	result
"> | a | b |\n> |---|---|\n>"	IndexError
"> | a | b |\n> |---|---|\n> " (marker + space)	IndexError
"> > | a | b |\n> > |---|---|\n> >" (nested)	IndexError
"> | a | b |\n> |---|---|\n> | 1 | 2 |\n>" (body row, then bare marker)	IndexError
"> | a | b |\n> |---|---|\n>\n" (trailing newline)	ok
"> | a | b |\n> |---|---|\n> | 1 | 2 |" (complete)	ok
"| a | b |\n|---|---|" (no blockquote)	ok
"> text\n>" (no table)	ok

A trailing newline makes it disappear, which is why this is easy to miss in a document renderer and
hard to avoid in an incremental one. I hit it rendering a partially-streamed LLM reply with
rich.markdown — > |---|---|\n> is not markdown anyone writes, but it exists for a few
milliseconds every time a quoted table is streamed a token at a time. rich parses in
Markdown.__init__, so it raises at construction and takes down the caller.

Suggested fix

The guard the sibling rules already use:

--- a/markdown_it/rules_block/html_block.py
+++ b/markdown_it/rules_block/html_block.py
-    if state.src[pos] != "<":
-        return False
+    try:
+        if state.src[pos] != "<":
+            return False
+    except IndexError:
+        return False

--- a/markdown_it/rules_block/heading.py
+++ b/markdown_it/rules_block/heading.py
-    ch: str | None = state.src[pos]
-
-    if ch != "#" or pos >= maximum:
+    if pos >= maximum or state.src[pos] != "#":
         return False

Happy to open a PR with these plus tests if the approach looks right — though if #190 is still the
preferred direction, that would fix this class of bug at the source and these two rules are simply
further evidence for it.

List your environment

• markdown-it-py 4.2.0 (the two lines above are unchanged on master as of this writing)
• Python 3.12, Linux
• No optional dependencies needed to reproduce