Refactor!!: Refactor the Annotator format to make it more extensible

[VaggelisD]

This PR refactors the annotation format by:

• Migrating the annotators to a new typing/ directory; This will help remove noise from the dialects/ files as we're making each dialect more type-aware
• Making the new format more expressive by moving from Expression -> Callable to Expression -> Dict which can carry additional state

[georgesittas]

Looks good, I'll eyeball the dialects locally as a sanity check. Let's benchmark this and then discuss what name to use instead of "expression spec".

[georgesittas]

/benchmark

[github-actions]

Benchmark Results

Legend:

• 🟢🟢 = 2x+ faster
• 🟢 = 1.1x - 2x faster
• ⚪ = No significant change (< 1.1x)
• 🔴 = 1.1x - 2x slower
• 🔴🔴 = 2x+ slower

Parsing Benchmark

Benchmark	bench_parse_main	bench_parse_pr
parse_sqlglot_tpch	5.48 ms	5.51 ms: ⚪ 1.00x slower
parse_sqlglot_crazy	15.3 ms	15.5 ms: ⚪ 1.01x slower
Geometric mean	(ref)	⚪ 1.00x slower

Benchmark hidden because not significant (6): parse_sqlglot_short, parse_sqlglot_long, parse_sqlglotrs_tpch, parse_sqlglotrs_short, parse_sqlglotrs_long, parse_sqlglotrs_crazy

---

Optimization Benchmark

Benchmark hidden because not significant (4): optimize_tpch, optimize_condition_10, optimize_condition_100, optimize_condition_1000