feat: Add snapshot functionality

[bitloi]

• Add snapshot() method to AgentFS SDK using checkpoint + file copy
• Add 'agentfs snapshot' CLI command for manual snapshots
• Add automatic snapshot on nested 'agentfs run' (detected via AGENTFS_SESSION env var)
• Add Snapshot error variant
• Add tests for snapshot functionality

closes #57, #232

[bitloi]

running 2 tests
test tests::test_snapshot_is_independent ... ok
test tests::test_snapshot ... ok

test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 100 filtered out; finished in 0.05s

[bitloi]

Hey @penberg, can you please check the pr?

[penberg]

Hey @bitloi! We had someone do this earlier #119 and I commented that I don't think the snapshotting logic is safe against concurrent writes.

[bitloi]

Hi @penberg, thanks for the feedback!

You're right - the current implementation has a race window between the WAL checkpoint and file copy where concurrent writes could slip through. I noted this limitation when implementing it.

A few questions:

1. Is the turso Connection::snapshot() API still planned? If so, I can update this PR to use it once available.

2. For the auto-snapshot on nested run (Issue Automatic snapshot on nested agentfs run #232), the snapshot happens in the same process before spawning the nested session, so there shouldn't be concurrent writers at that moment. Would that specific use case be acceptable even with the current approach?