fix: log Comet execution errors before Spark can obscure them

[andygrove]

Which issue does this PR close?

Closes #.

Rationale for this change

When a NullPointerException or AssertionError occurs in Comet execution code, Spark's QueryExecution.toInternalError catches it and wraps it in a generic [INTERNAL_ERROR] The "collect" action failed. You hit a bug in Spark or the Spark plugins you use. message. This hides the original Comet context (which operator failed, what the root cause was), making it very difficult for users to debug or report issues.

What changes are included in this PR?

Adds error logging in the CometPlan trait using the template method pattern:

• CometPlan now overrides doExecuteColumnar, doExecute, and doExecuteBroadcast with a withErrorLogging wrapper that catches exceptions, logs them with Comet-specific context (operator class name and method), and re-throws the original exception.
• All Comet operators now override doExecuteCometColumnar, doExecuteComet, and doExecuteCometBroadcast instead of the Spark methods directly.
• The original exception is preserved and re-thrown unchanged, but the Comet-specific context is captured in the logs before Spark has a chance to obscure it.

How are these changes tested?

Verified manually by injecting a NullPointerException in CometNativeExec.doExecuteCometColumnar and running CometExecSuite. Confirmed that the error is logged with the Comet operator name and original cause before Spark wraps it.