codegen: CRLF and lone CR line endings (#230, line-ending part) - #252

Normalize CRLF and lone CR code/comment line endings in Lexer initialization after checking the original source size. Construct that lexer at the front of _parse_bounded and pass lexer.source to pragma extraction and Parser, so transpile, transpile_full, the request inventory API, and JSON/Pyodide glue use the same text and locations. Keeping the rule in Lexer also covers direct token users and expression reparsing.

Preserve quoted string contents, including raw CR characters, escaped delimiters and triple-quoted strings. LF source passes through unchanged. Add an Unreleased changelog entry and update the matching project memory.

Regression tests are in the preceding tests-only commit. No tests or builds were run on this Mac, as instructed; validation here was static Python syntax parsing and git diff checks only.

Signed-off-by: luisleo526 <[email protected]>

A CRLF pair is a line terminator wherever it appears, including inside a
multiline string, a wrapped one-quote string and right after a backslash, so
a script saved with CRLF now yields the same string values, C++ and
diagnostics as the LF form. A lone CR inside a string stays data (the
committed regression tests pin it); lone-CR-only files with a multiline or
wrapped string are a documented limit.

Also build the pragma pre-pass lexer inside its try, so a whole-source size
error keeps the recovery it had before the size check moved into the lexer.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…#230)

A global replace of CRLF turned "\r\r\n" (a lone CR followed by a CRLF, the
artifact of a file converted twice) into CR+LF, which the scanner then read
as one break where the LF reading has two: a blank line was lost, later
diagnostic line numbers were one short and a comment line merged with the
next. The CRLF pair is now folded only inside the scanner's string states
(and right after a backslash), and outside strings the existing branch takes
CRLF and a lone CR as one break each, in order. A CRLF inside a one-quote
string applies the same unwrapped-line recovery as the LF form, and that
recovery also treats a following CRLF blank line as blank.

Tests compare tokens, C++ and diagnostics with the LF form of the same
script for "\r\r\n" after code, after a comment and inside both string kinds.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Brings the branch up to date with codegen main (CHANGELOG backfill, #248). The
merge is clean: CHANGELOG.md auto-merged and keeps both main's entries and the
Unreleased note for CRLF and CR line endings (#230).

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

# Conflicts:
#	CHANGELOG.md