Allow %empty followed by %prec

[ratmice]

If one applies the following patch to the grammar in the zip archive linked from #473
to uncomment the %prec directives, and run the file through nimbleparse.

--- postgresql-16-naked.y       2024-11-21 10:36:08.000000000 -0800
+++ postgresql-16-naked.uncommented.y   2025-03-22 10:20:37.708582643 -0700
@@ -4572,7 +4572,7 @@

 opt_existing_window_name :
        ColId
-       | %empty //%prec Op /*16L*/
+       | %empty %prec Op /*16L*/
        ;

 opt_partition_clause :
@@ -4842,7 +4842,7 @@
        | WITH /*13N*/ UNIQUE /*11N*/
        | WITHOUT /*13N*/ UNIQUE /*11N*/ KEYS /*12N*/
        | WITHOUT /*13N*/ UNIQUE /*11N*/
-       | %empty //%prec KEYS /*12N*/
+       | %empty %prec KEYS /*12N*/
        ;

 json_name_and_value_list :

You currently get an error "%empty used in non-empty production".

[Error] in postgresql-16-naked.uncommented.y
    4575|       | %empty %prec Op /*16L*/
              %empty used in non-empty production

    554|        | MODE_PLPGSQL_EXPR PLpgSQL_Expr
                              ^^^^^^^^^^^^ Unknown reference to rule 'PLpgSQL_Expr'

This relaxes that error message so that %prec is accepted after %empty.
in #453 there was also a question about whether %prec x should be accepted without quotes around 'x'.
I believe that the usage of %prec Op here seems to confirm that.

I wasn't able to test this with the provided input.txt, the lexer seems to be missing the necessary tokens.
but with this change I was able to run the modified grammar over the empty string. Producing an empty parse tree.

echo -n '' | ./target/release/nimbleparse -y original postgresql-16-naked.l postgresql-16-naked.uncommented.y -
[Warning] in postgresql-16-naked.uncommented.y
    3686|       select_no_parens %prec UMINUS /*22R*/
                                  ^^^^^^ Unused token


parse_toplevel
 stmtmulti
  toplevel_stmt
   stmt

[ltratt]

I think this is an improvement, even though I'm sure we could improve things further again.

[ratmice]

Needed a rustfmt (Should I just automatically squash for changes like this?).

[ltratt]

Yes for small changes like this, please feel free to squash directly (but don't rebase because the makes the compare button in github useless).

[ratmice]

Squashed.