Summary

Columns with dots (or other special characters) in their names fail when used in VISUALIZE mappings with quoted identifiers — e.g. "variable.dotted" AS y. The parser was storing the raw source text including surrounding double-quotes, so validation couldn't match them against the actual (unquoted) Arrow schema column names.

• Promotes the private unquote helper from stat_aggregate.rs to naming::unquote_ident (inverse of quote_ident)
• Applies it at parse time in builder.rs for both explicit and implicit column mappings

Fixes posit-dev/ggsql-python#3

Repro

duckdb reproduce.db -c "CREATE TABLE test AS SELECT 1 AS numbers, 3 AS \"variable.dotted\" UNION ALL SELECT 2, 4"
ggsql exec --reader "duckdb://reproduce.db" \
  'SELECT * FROM test VISUALIZE numbers AS x, "variable.dotted" AS y DRAW line'

Before: Validation error: Layer 1: aesthetic 'pos2' references non-existent column '"variable.dotted"'
After: renders successfully

Test plan

• New unit tests for unquote_ident (basic cases + quote/unquote roundtrip)
• All existing tests pass (cargo test --lib naming — 32 passed)
• End-to-end CLI verification with dotted column name