[baylanger]

Add python -m pyhik check

A check to run by hand against a real camera or NVR. Unit tests have to run
without hardware. This is for people with a device who want to try the
image settings code or report a problem.

python -m pyhik <command> is a small dispatcher (pyhik/__main__.py) that imports a
command only when it is asked for. check is the first command, so others (for example
dump) can be added later.

Run

export CAMERA_HOST=192.168.1.100
export CAMERA_USERNAME=admin
export CAMERA_PASSWORD='passwd123'
python -m pyhik check

Optional: CAMERA_PORT (80, or 443 with HTTPS), CAMERA_SSL=1, CAMERA_VERIFY_SSL=0 for a
self-signed certificate. --host, --port, --username, --ssl and --no-verify do the same as
CAMERA_HOST, CAMERA_PORT, CAMERA_USERNAME, CAMERA_SSL=1 and CAMERA_VERIFY_SSL=0, and win over
them. The password is read only from the environment, never from the command line, and never printed.

What it does

Command	Effect
python -m pyhik check	Read-only. Lists channels and every image setting with its path and value.
... --write	Changes Color/brightnessLevel by one step, reads it back, restores it.
... --write --path Color/contrastLevel	Same, for the setting you choose: take its path from the read-only list. A number moves one step, a switch flips (true to false or false to true) and a select takes the first option that is not the current one. --path is only used with --write.
... --conflicts	WDR / HLC / BLC / defog conflict matrix.
... --bad-password	A wrong password, tried through ISAPIClient and HikCamera: at least two failed logins.
... --all	All of the above.
... -v	Also shows pyhik's debug log (the retries).

A plain --write ends by naming a few switch and select paths to try with --path. If a change
switches another setting off to make room (WDR, BLC, HLC and defog exclude each other), that setting
is put back too and listed.

One channel is checked per run (the first listed unless --channel is
given). On an NVR or DVR the attached camera is used through the ImageProxy
endpoints:

python -m pyhik check --channel 3 --write --path Color/contrastLevel
python -m pyhik check --channel 1 --write --path enableImageLossDetection

The NVR is detected from the device type. That is not enough on its own: some NVRs report
IPC, and the device information may not be readable at all. When the camera image endpoints fail
and the NVR ones work, the NVR endpoints are used, and it prints which it chose and why.
--proxy / --no-proxy override the detection.

The device information is asked once per run, whether it answers or not. When it fails, the device
line shows the error with its HTTP status (for example
model=? (ISAPIError: Request failed with status 503)) and the run goes on.

Safety

• Everything that changes a value puts the original back afterwards, even
  after an error, and prints NOT RESTORED, set it back by hand if it
  could not.
• It stops at the first authentication error, since repeated failures can
  lock an account. --bad-password makes at least two failed logins (one per
  client, more if a client retries); do not use it on an account that is close
  to being locked.

Tested

Unit tests (test/test_check.py) cover the NVR detection, the write path (number, switch, select, a conflict, a failed restore), the connection options (host, port, username, HTTPS, certificate check) and the python -m pyhik dispatch, with fake devices.
On real camera DS-2CD2385G1
On a DS-7608NI-EV2/8P NVR (V3.4.96): 8 channels listed, a read, and a write of a switch
(enableImageLossDetection) through ImageProxy with the read back and the restore both
checked.

Dependencies

[ ] Requires #151 (PR4 List image channels and support NVR-attached cameras through ImageProxy) to be merged first.

[mezz64]

I'm onboard with adding this, but considering the ISAPI functionality is only a subset of what pyHik can do, I don't know that the path should be simply called "check" as i'd expect that to check the basic event stream functionality. I think this is better suited as "checkiaspi" or something similar so a separate module can be added later to perform the same function on the base event stream monitoring code.

Also, the readme changes need cleaned up now that #151 is merged and I'd prefer you split out the xmldict change into a separate PR.

Otherwise a review flagged a couple small issues:

• An error during the restore step in check_write can hide the original error, and the restore keeps sending requests after an auth error.
• The restore writes back every value that changed, not just the ones in result.also_changed.
• TESTING.md says cd pyhik; it should be cd pyHik.