Replace Python regex-based string/comment stripper with SBCL's actual
reader. For each lisp block, feeds the code to read-from-string in a
loop (reading all forms). Correctly classifies: package errors (not a
paren problem), reader errors (extra/missing closes), EOF (missing
closes Handles all Common Lisp reader edge cases: character literals
(#\( #\) #\;), block comments, string escaping
Scans #+begin_src lisp ... #+end_src blocks, strips strings and
comments, reports unbalanced parens per-block with line numbers.
Detects unterminated blocks (no matching #+end_src).
Zero dependencies (stdlib Python). Called from AGENTS.md step:
projects/check-parens/check-parens org/file.org
