Decode cpu-clock samples as stacktraces

[jeet-dekivadia]

Summary

• route cpu-clock perf events through the existing sampled stacktrace decoder
• add expect coverage using a reported sampling-fallback frame

Why

When Intel PT is unavailable, perf records sampling fallback events as cpu-clock. Their payload has the same callstack shape as cycles, but the event-header decoder rejected the name before stacktrace parsing could run.

Validation

• git diff --check
• exercised the event-header pattern against the reported cpu-clock:uH line
• added expect coverage for the sampled callstack path

Fixes #283
Fixes #311

[jeet-dekivadia]

I rechecked this today while it is waiting for review. The patch routes perf sampling fallback events named cpu-clock through the same stacktrace decoding path as the normal sampled events, instead of rejecting the header because the event name is not one of the hardware counters.

The regression is intentionally tied to the reported cpu-clock:uH header shape and keeps the callstack parsing behavior shared with the existing stacktrace decoder. DCO is green, and the PR has no red checks.