Xiaoya refactor publisher

[xiaoyachong]

The main purpose of this PR is to move the Tiled publisher from LSE to arroyoXPS, and to have arroyoXPS send results to LSE via ZMQ instead of using a WebSocket connection.

1. Results are written to Tiled and a ZMQ message is sent to LSE using the same publisher, XPSTiledResultPublisher.
2. In the XPSTiledResultPublisher, we import:
   from arroyosas.schemas import RawFrameEvent, SASStart, SASStop, SerializableNumpyArrayModel
   Since both arroyosas and arroyoxps send the same message types (RawFrameEvent, SASStart, SASStop) to LSE, it may make sense to move these schemas into arroyopy and rename them (e.g., remove the SAS prefix).
3. To handle mid-scan issues, I avoid writing to Tiled unless a start message has been received. For example, if splash_timepix starts before arroyoxps, the Tiled publisher logs the following message:
   Received scan 'None' frame 49 but no start message has been received yet — discarding frame to avoid Tiled index desync.
4. The Tiled structure:

{tiled_base_uri}/{tiled_prefix}/
│
├── {root_segments1:lse_live_results}/
│   └── {YYYY}/
│       └── {MM}/
│           └── {DD}/
│               └── {experiment_name}/
│                   └── {UUID}/
│                       └── feature_vectors  
│                           (table storing feature vector results)
│
└── {root_segments2:xps_processed_images}/
    └── {YYYY}/
        └── {MM}/
            └── {DD}/
                └── {UUID}/
                    └── xps_averaged_heatmaps  
                        (3D array storing shot-mean heatmaps)

Companion PRs:
LSE: mlexchange/mlex_latent_explorer#79
arroyosas: als-computing/arroyosas#22

[dylanmcreynolds]

I think we can make this simpler. See commend.

[dylanmcreynolds]

I approved. Is this mergable?

[taxe10]

All looks good. I left some very minor comments.

[taxe10]

I think ARRAY_KEY is referenced before it's defined

[xiaoyachong]

It's a module-level variable defined as "xps_averaged_heatmaps".

[taxe10]

Suggested change

	self.seen_scans_file = seen_scans_file
	self.seen_scans_file = Path(seen_scans_file)

[xiaoyachong]

The seen scans are stored as UUIDs, so the path is not used.

[taxe10]

Suggested change

	seen_scans_file=seen_scans_file,
	self.seen_scans_file = Path(seen_scans_file)

[xiaoyachong]

The seen scans are stored as UUIDs, so the path is not used.

[taxe10]

Should we remove this comment?

[xiaoyachong]

Yes. I just removed it.