Fix stale frame returned by capture API on rapid successive calls
> [!NOTE]
> Before submitting code changes...
> * Please do note that this is a GPL v3.0 open source project.
> * Please read the
[CONTRIBUTING](https://github.com/spice2x/spice2x.github.io/blob/main/CONTRIBUTING.md)
guide.
> * Maintainers reserve the right to reject or modify your submission
without reason.
> * No new compiler warnings must be introduced. Check the CI build
results.
>
> Feel free to remove this section after you have read it.
## Link to GitHub Issue or related Pull Request, if one exists
https://github.com/spice2x/spice2x.github.io/issues/582
## Description of change
When using the SpiceAPI `capture` endpoint for taking screenshots, if
you call multiple times in a row in a short period, there's a good
chance for duplicate / stale data to be returned, and you end up with
multiple identical screenshots. This is due to a race condition in the
graphics capture pipeline because the capture data is never reset to
null between invocations.
`graphics_capture_receive_jpeg()` never resets the capture buffer's data
pointer to `nullptr` after consuming it. The condition variable wait
predicate checks `data != nullptr`, so subsequent calls find the
predicate already satisfied and return immediately with the previous
frame instead of waiting for the render thread to produce a fresh one.
Fix: null out `capture.data` after copying the shared_ptr (which keeps
the pixel data alive via refcount) but before releasing the lock.
## Testing
Tested this locally and invoked the capture API 10+ times in a row
without any duplicates. Previously, every other invocation would be a
duplicate, basically.
