Updates to digitiser aggregator, and adding fields to spans

[Modularius]

Summary of changes

cache.rs

1. Swapped VecDeque to HashMap<FrameMetadata,_> and implemented resulting changes
2. Added metadata to push function instrumentation.
3. Added get_num_partial_frames function.
4. Added test_metadata_equality test.
5. push now returns &impl SpannedAggregator, and dependence on FindSpan(Mut) removed.

partial.rs

1. Added metadata to Frame span.

main.rs

1. Refactored on_message to process_kafka_message and process_digitiser_event_list_message functions.
2. Added metadata and other fields to process_kafka_message and process_digitiser_event_list_message instrumentation
3. Instrumented root_as_digitizer_event_list_message by wrapping with span_of_root_as_digitizer_event_list_message

frame_metadata.rs

1. Implemented Hash trait for FrameMetadata
2. Implemented PartialEq for FrameMetadata as a temporary measure to exclude veto_flags (this will be removed later).

Instruction for review/testing

1. Do fields added to spans conform with the rules set out in tracing.md?
2. General code review.

[DanNixon]

Swapped VecDeque to HashMap<FrameMetadata,_> and implemented resulting changes

Why was this done?

[Modularius]

It was the cleanest way to make the push function return a pointer to the PartialFrame object, plus it seems a more natural container to use for matching on FrameMetadata.

[DanNixon]

The issue is that in doing so you remove the deterministic ordering of frames emitted from the aggregator in the case where multiple frames are held in the cache.

[Modularius]

I see, perhaps we can keep a VecDeque with the FrameMetadata so they can be emitted in sequence. I'll have another look at reimplementing the VecDeque solution first though.

[DanNixon]

I think I am missing what was wrong with using VedDeque as it was.

[Modularius]

I was tripping over the borrow checker trying to return a ref to its contents, whilst also possibly modifying it if a new PartialFrame was needed

[DanNixon]

I was tripping over the borrow checker trying to return a ref to its contents, whilst also possibly modifying it if a new PartialFrame was needed

I assume the only reason you need to return anything from push() is to handle span linking here?
I think this makes more sense to be handled in push() itself.

[Modularius]

That's right. I prefer to return from push to decouple as much of the Otel handling from the rest of the functionality by keeping it in main.rs.

I managed to find a way to reimplement VecDeque though.

[DanNixon]

👍