Add Breakpoint Label frame to optimize debug stepping performance

[JustinGrote]

PR Summary

Implements DAP DelayedStackTraceLoading and returns an artificial breakpoint label frame based on the invocation info ASAP once the debugger stops. The rest of the stack trace is then returned on a future request that doesn't block the client UI.

Capture.mp4

PR Context

Debug stepping takes 300ms+ due to waiting for stack trace collection via the slow Get-PSCallStack "hack". Further improvements will come around this later.

[JustinGrote]

@SeeminglyScience @andyleejordan I made this a fairly minimal PR to be easy to review. Additional things I plan to do next as future incrementals

• Add start/end/column positioning to all stack frames (it's available in the position property, just has to be serialized properly)
• Separate variable and frame processing into independent resolution paths
• Implement "fast path" for local runspace sessions to use the runspace debugger properties and methods rather than going thru the pipeline to fetch the frame info, ~2-3ms vs 300ms-800ms
• Move all the scope/stack/variable resolution to a new async-based DebugStoppedContext class that can be disposed and re-instantiated cleanly and kept separate from the DebugService concerns, since LSP spec says no IDs/etc. should be reused between debug adapter stop/resumes

[andyleejordan]

@SeeminglyScience @andyleejordan I made this a fairly minimal PR to be easy to review. Additional things I plan to do next as future incrementals

• Add start/end/column positioning to all stack frames (it's available in the position property, just has to be serialized properly)
• Separate variable and frame processing into independent resolution paths
• Implement "fast path" for local runspace sessions to use the runspace debugger properties and methods rather than going thru the pipeline to fetch the frame info, ~2-3ms vs 300ms-800ms
• Move all the scope/stack/variable resolution to a new async-based DebugStoppedContext class that can be disposed and re-instantiated cleanly and kept separate from the DebugService concerns, since LSP spec says no IDs/etc. should be reused between debug adapter stop/resumes

I'm super excited for all of this.

[andyleejordan]

@SeeminglyScience and I are going to take a closer look at this just to be sure nothing got missed.

[JustinGrote]

This has been rebased to the latest prerelease and is good for re-review.

[SeeminglyScience]

Thanks Justin! Couple of questions and minor requests

[andyleejordan]

Loving this, we finally got to review through it. Just a couple changes requested and then we want to get into pre-release.

[andyleejordan]

Suggested change

	//We have an artificial breakpoint label, so just copy the stacktrace from the first stack entry for this.
	// We have an artificial breakpoint label, so just copy the stacktrace from the first stack entry for this.

Could ou explain why this is either 0, or (and this is what's changed) FrameId - 1? That means there are two instances where we get scope 0: FrameId == 0 and FrameId == 1.

[JustinGrote]

Uh, there was a reason, but I honestly don't remember at this point. Let me check

Copilot reviewed 5 out of 5 changed files in this pull request and generated no suggestions.

[JustinGrote]

@andyleejordan @SeeminglyScience your opinions have officially been deemed irrelevant by our AI overlords, please approve as is /s