feat: Wire DataFusion function Claude skill and csc

[comphead]

Which issue does this PR close?

Part #4150

Rationale for this change

Spark csc(x) = 1 / sin(x) was unsupported in Comet, falling back to Spark. The upstream datafusion-spark crate (53.1.0) ships a Spark-compatible SparkCsc implementation, so we can wire it
directly with no native Rust written in this repo.

This PR also adds a new Claude Code skill that codifies the wiring workflow used here, so future contributors can apply the same recipe consistently.

What changes are included in this PR?

Wire Spark Csc to datafusion-spark:

• spark/src/main/scala/org/apache/comet/serde/QueryPlanSerde.scala — register classOf[Csc] -> CometScalarFunction("csc") in mathExpressions.
• native/core/src/execution/jni_api.rs — register SparkCsc in register_datafusion_spark_function.
• docs/source/user-guide/latest/expressions.md — add Csc to the Math Expressions table.
• docs/source/contributor-guide/spark_expressions_support.md — flip csc checkbox to supported.

Upstream function: datafusion_spark::function::math::trigonometry::SparkCsc. Wiring follows Pattern B (one-line Scala passthrough + one-line UDF registration).

New Claude Code skill wire-datafusion-function:

• .claude/skills/wire-datafusion-function/SKILL.md — project-level skill that walks contributors through wiring an existing DataFusion or datafusion-spark function into Comet. Documents the three
  wiring patterns (passthrough, UDF registration, custom serde with restrictions), the SQL-test coverage requirements (parquet-backed inputs, NULLs, FP edge cases, integer/decimal min/max), and the
  hand-curated docs that must be updated. Complements the existing implement-comet-expression skill — used here to scaffold the csc wiring.