v0.3.0

fix: bind upload sessions to their target, keep pending scans in sync

- Resumable sessions record file id, parent, name, precondition and content
  hash; a session is resumed only for the exact plan it was opened for, and
  the returned identity is verified before success is recorded
- Pending folder scans follow renames and removals, resolve their path by id
  when they run, and a vacated path queues a shallow re-list of its parent so
  a suppressed duplicate is rediscovered
- Subtree pushes verify every cached ancestor before creating a folder
- Session persistence errors fail the upload instead of being ignored
- Regression suite against a loopback Drive fixture

Signed-off-by: Rohit Yadav <rohit@scaleninja.com>

v0.2.0

drive: case around upload stall

The upload stall budget now assumes a 100 KB/s floor instead of 200
KB/s. On a slow uplink the previous value could exhaust the six
in-process attempts on a multi-gigabyte file, even though each attempt
resumed from the last byte. The cost is a longer wait on a genuine
stall, which keepalive already catches when the peer is dead.

Signed-off-by: Rohit Yadav <rohit@scaleninja.com>

v0.1.0

drivesync: make it fast and accurate

Add documentation in README and also how it works, how to use it.

Signed-off-by: Rohit Yadav <rohit@scaleninja.com>