Consider Removing New Line Characters when Comparing Computed Constructs

I'm working with a Computed Column construct, which is really long and therefore contains line breaks for readability.

When this construct gets imported into the underlying database (Postgres in my case), the construct gets reformatted with line breaks inserted in different places.

This creates a small side-effect, where alembic now thinks that the original computed clause provided by the migration file is different to the one in the database, and therefore emits a warning.

/usr/local/lib/python3.10/site-packages/alembic/autogenerate/compare.py:1090: 
UserWarning: Computed default on ***  cannot be modified
util.warn("Computed default on %s.%s cannot be modified" % (tname, cname))

After inspecting the code, I found this method, which appears to sanitize the database computed clause as well as the local one:

https://github.com/sqlalchemy/alembic/blob/fc451e60731ad5b8d73255f0a1996e1d4e0f20de/alembic/autogenerate/compare.py#L1047-L1055

I feel that we should include line breaks in the list of tokens that should be removed before a comparison takes place, but happy to hear what thoughts you have on this.

Thanks!

Start in alembic/autogenerate/compare.py around lines 1047-1055, where the database and local computed clauses are sanitized before comparison. Reproduce the issue with a multiline Computed Column against PostgreSQL, then verify that equivalent clauses with differently placed line breaks no longer produce the computed-default warning.