Apply PIVOT aliases before schema validation

[osipovartem]

Which issue does this PR close?

Closes a Snowflake compatibility gap where chained PIVOT clauses produce duplicate intermediate literal column names before the outer table alias can rename them.

Rationale for this change

Snowflake supports assigning a complete column alias list to the result of chained PIVOT clauses. DataFusion currently builds the second aggregate first, rejects its duplicate output names, and never reaches apply_table_alias.

What changes are included in this PR?

• normalize and apply explicit PIVOT output aliases while constructing group-by and aggregate expressions
• preserve the existing alias-count validation and final table alias
• add an integration test with two chained PIVOT clauses whose implicit literal names collide

Are there any user-facing changes?

Queries such as ... PIVOT(...) PIVOT(...) AS p(col1, col2, ...) now plan and execute when the explicit aliases make the final schema unambiguous.

How was this tested?

• cargo +stable test -p datafusion-sql --lib --offline: 92 passed
• cargo +stable test -p datafusion --test core_integration pivot_unpivot --offline: 3 passed
• cargo +stable clippy -p datafusion-sql --all-targets --offline -- -D warnings
• cargo +stable fmt --all --check

Stacked on #47, which contains the DF55 Snowflake compatibility patches.