perf: skip redundant lookahead simulations in Condition dispatch and expression lists

[fudianchn]

AI disclosure: this change was prepared with AI coding agents.

What

Skip two redundant lookahead simulations on the hottest parse paths: the deep jj_2 phase simulation guarding the main branch of Condition() / SQLCondition(), and the Oracle/PostgresNamedFunctionParameter probes in ComplexExpressionList().

Why

The main branch of Condition() / SQLCondition() is the last alternative of its choice, so its lookahead can only accept or reject what the productions themselves accept or reject immediately after: entering it directly and letting them fail is equivalent, without re-walking the whole expression tree first. Under JFR profiling of parseSQLStatements this simulation collected about 17% of the execution samples, paid by every plain condition on top of the real parse.

The named-parameter probes can never match a dotted name (f(a.b => 1) fails to parse on master with or without this change), so a check of the second token decides them exactly like the 2-token scan did.

How

• Condition() / SQLCondition(): gate the main alternative with LOOKAHEAD({ !interrupted }) instead of the implicit deep simulation; the EXISTS alternatives ahead of it and every inner lookahead are untouched.
• ComplexExpressionList(): gate the four named-parameter alternatives with isNamedParameterAssignmentAhead(), a two-kind check on getToken(2).

Root cause

Not applicable (performance).

Testing

• 26 probe statements covering the EXISTS forms (identifier usage, EXISTS (subquery), exists.col, exists (1), ternary ? EXISTS :arr :), named parameters (a => 1, a := 1, OUTER => 1) and dotted-name rejections: parse results and failures identical to master.
• Full suite: 10,344 tests, the same 35 environment failures as master on my machine (33 Mockito initialization, 2 file access), zero new failures; AST dump of performance.sql byte-identical to master (54 statements).
• Benchmark JSQLParserBenchmark.parseSQLStatements (JDK 17, 1 fork, 3x10s warmup, 10x1s, jars interleaved): 21.32-21.40 vs 24.58-24.83 ms/op against master over 3 pairs, error intervals not overlapping (-13% to -14%).
• Not covered locally: the Mockito-based interruption test cannot run in my environment (it fails on master too); the !interrupted gate follows the existing LOOKAHEAD(n, { !interrupted }) convention used across the grammar. Malformed conditions can now fail inside the productions instead of at the choice dispatch; the failure class is unchanged.

Verification of the original issue

Not applicable (no issue).