fix(sqlite): transpile Postgres escaped strings as text literals

[russellromney]

Follow-up to #7677. Postgres E'...' strings parse as ByteString, and SQLite currently has no byte string delimiter, so generation falls back to an unsupported empty string. That drops valid text payloads during Postgres -> SQLite transpilation.

This makes the SQLite generator render ByteString as a normal quoted text literal, preserving the decoded payload.

Examples fixed:

SELECT E'a\nb'         -- SELECT 'a
b'
SELECT E'a\tb'         -- SELECT 'a	b'
SELECT E'a\'b'         -- SELECT 'a''b'
SELECT LENGTH(E'a\nb') -- SELECT LENGTH('a
b')

Checks run:

uv run pytest tests/dialects/test_postgres.py -q -k postgres
uv run pytest tests/dialects/test_sqlite.py -q
uv run ruff check sqlglot/generators/sqlite.py tests/dialects/test_postgres.py
uv run ruff format --check sqlglot/generators/sqlite.py tests/dialects/test_postgres.py

[geooo109]

@russellromney thanks for the PR.

The PR doesn't solve the problem in general right ? it misses cases like r"SELECT e'\x41', length(e'\x41')", SELECT e'\\\x41', length(e'\\\x41'), right ? I'm not sure if they can be solved, because they can also be values of column, so the solution should be compex.

[georgesittas]

It appears that solving the general case would entail pinning down every possible escaped byte sequence and de-escaping it by emitting the actual character. I'm not sure to what extent this is possible or what the end solution will look like, but it seems too complicated.

I'd rather we don't do this right now and call unsupported, which should happen already.

[georgesittas]

Closing this for now, but happy to continue discussing.