Parser: fix exponential parse time on compound keyword chains

[LucaCappelletti94]

Same family as #2344 and #2349. A reserved keyword in field position (e.g. NOT in .not-b.not-b...) drives parse_prefix -> parse_not -> parse_subexpr, re-walking the chain at every segment and doubling the work. The result is always rejected (UnaryOp), so the speculation is wasted, and parse_identifier in the existing None branch produces the same Identifier directly. The solution I identified is to skip the speculative parse_prefix in parse_compound_expr when the next token is a Word not followed by (. Surfaced by subql fuzzing on a 527 B input that took 2.65 s and now takes 270 us.

N	before	after
5	63.3 us	8.06 us
10	1.97 ms	15.1 us
15	65.3 ms	22.4 us