Add simplified API v2 with single structure endpoint (#36)

[Titi211179]

Addresses #36

This PR implements a simplified API v2 that reduces the number of
sequential calls needed to discover and access data in a file.

Current v1 flow (3+ calls)

1. GET /api/v1/blockdata/ → block list + segment counts
2. GET /api/v1/segmentdata/ → signal metadata for one segment
3. GET /api/v1/analogsignaldata/ → actual signal data

Each call re-opens the file. Users must discover structure step by step.

New v2 flow (2 calls)

1. GET /api/v2/structure/ → complete file hierarchy (blocks, segments,
   signal metadata, spike train metadata) in a single response
2. GET /api/v2/data/analogsignal/ or /api/v2/data/spiketrains/ →
   actual data

Changes

• New v2 data models: FileStructure, BlockStructure,
  SegmentStructure, SignalInfo, SpikeTrainInfo — rich metadata
  without payload data (sampling rate, duration, channel count,
  sample count)
• New api/resources/v2.py with 3 endpoints: /structure/,
  /data/analogsignal/, /data/spiketrains/
• v2 mounted at /api/v2/ prefix
• v1 endpoints remain untouched for backward compatibility
• Includes refresh_cache parameter from Add an option to the API to ignore/refresh the file cache #40
• Clearer error messages with actual index values
• Version bumped to 2.0

Example

GET /api/v2/structure/?url=https://example.com/data.nwb

Returns all blocks → segments → signal metadata in one response,
so the client knows exactly what to request next.