Propagate cancellation causes through task trees

[infiton]

Propagate cancellation causes through Node, Task, and Scheduler cancellation traversal so bulk cancellation reuses the originating Async::Cancel::Cause instead of regenerating a fresh cause/backtrace for each descendant task.

This reduces redundant Cancel::Cause.for / backtrace capture work during scheduler-driven bulk cancellation and shutdown. Descendant tasks now observe the originating cancellation cause rather than a per-node synthetic cause generated during traversal.

A benchmark was added in benchmark/cancel_cause.rb to demonstrate the improvement against main:

# TASKS=1365
main:   cause_for_calls=1365 allocated_objects=26009 duration_ms=7.42
branch: cause_for_calls=1    allocated_objects=5545  duration_ms=5.79

# TASKS=5461
main:   cause_for_calls=5461 allocated_objects=103833 duration_ms=35.03
branch: cause_for_calls=1    allocated_objects=21929  duration_ms=23.94

Types of Changes

• Performance improvement.

Contribution

• I added tests for my changes.
• I tested my changes locally.
• I agree to the Developer's Certificate of Origin 1.1.

[ioquatix]

Thanks for your contribution. The only change I made was the state transition methods (ending with bangs) take positional arguments to be consistent with other state transition methods.

FYI, from my point of view, it's good to do less allocations during cancellation so this is a good idea. In general, error handling (which usually involves cancellation) should avoid doing unnecessary work. Otherwise you can have cascading failures (e.g. run out of memory).