Implement Spark SQL parser extensions and enhance analytic functions

[damccorm]

This PR brings in several enhancements for analytic/window functions and adds Spark-compatible SQL parser extensions:

• Analytic Functions:
  • Added support for CUME_DIST().
  • Implemented IGNORE NULLS routing for FIRST_VALUE and LAST_VALUE (needed for pandas ffill/bfill compatibility).
  • Switched to a ValueHolder accumulator in these combiners to safely handle null values without throwing NPEs.
  • Fixed BeamWindowRel to support functions with no input columns (like COUNT(*)).
• Planner/Parser Fixes:
  • Fixed a Volcano planning failure in BeamWindowRule by explicitly converting the input to BEAM_LOGICAL convention (mirrors what we do in Sort/Aggregation rules).
  • Added Spark-style parser extensions to parserImpls.ftl and config.fmpp: support for angle-bracket types (e.g., array<int>, map<string, int>), type constructors (e.g., float(x)), and the :: infix cast operator.
• Spark-style Collection Constructors:
  • Registered ARRAY(...) and MAP(...) function-form constructors in CalciteQueryPlanner so they validate correctly (bracket form ARRAY[...] is standard, but function form is Spark-specific).
  • Added tests in BeamSqlDslArrayTest and BeamSqlMapTest.
• Intervals: Added NanosDuration decoding to BeamTableUtils to correctly handle Calcite's interval representations.

Added tests to BeamAnalyticFunctionsTest for CUME_DIST and COUNT(*) OVER ().