feat: add telemetry to the ingestion pipeline

[nordfjord]

Adds some basic telemetry to the ingestion pipeline.

All attributes are prefixed with propulsion.

Attribute	Description
category	The category being processed
stream_name	The full stream name being processed
stream_id	The id of the stream being processed
batch_size	The size of the batch being processed
first_timestamp	The receive timestamp of the first event in the batch being handled
lead_time_ms	The lead time for processing the batch

The above screenshot shows a run of the It doesn't read the tail event again test. It appends 20 events via an equinox decider before starting to poll. You can see it create the checkpoints table, read from it, poll for events, and then fetching two batches before processing a stream batch. Then you see the tailSleepInterval kick in with a wait for 1s before storing the checkpoint and ending the test.

This gives users the opportunity to ship these spans to various observability providers and ask questions like "What's my average lead time for processing category X"