Filewriter: implement structure

Implement the initial top-level structure of the filewriter process, to prove that we have a design that will work in practice. This should follow DS ADR 10.

Should include:

• Trait definitions for writer modules
• Kafka-listening infrastructure (including pooled architecture)
• HDF5-writing infrastructure
  • Including writing fully-static datasets/groups
• Infrastructure for parsing dynamically-provided nexus_structure (part of run start)
  • Including configuration parameters for each writer module
• Logging
• Metrics
• Testing
  • Including unit tests for individual writer modules
• Benchmarking
  • Include a bench for each writer module
  • Include one overall top-level bench for writing a small but representative file in-memory
• Docs
• Error propagation with appropriate context for debugging (likely miette + thiserror)
• Two different writer modules, which may write simple/incomplete data at this stage.
  • E.g. simplified IXseblock and simplified NXevent_data

Functionally, the end point of this ticket is that we have a filewriter process that can listen to streams of data from Kafka and write some data from those streams, in a simplified/incomplete way, to a HDF5/NeXus file.