[m-rev.] for post-commit review: improve diagnostics for bad number literals

[Zoltan Somogyi]

For review by anyone. I would like feedback on two things
in particular: the text of the error messages, and whether
we should continue to generate an error for an underscore
before the exponent in a float. I see no reason for the ban,
and any motivation rooted in Prolog compatbility is long
in the past.

Zoltan.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Log.lit
Type: application/octet-stream
Size: 544 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20230619/f8f9ff9f/attachment-0002.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: DIFF.lit
Type: application/octet-stream
Size: 41773 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20230619/f8f9ff9f/attachment-0003.obj>