Use hub clock instead of frame clock for hub 0 devices

[bparks13]

This PR changes the devices on hub 0 to use the hub clock instead of the frame clock to save timestamps.

Devices modified:

• AnalogIO
• DigitalIO
• MemoryMonitor
• HarpSyncInput

Fixes #141

[jonnew]

@aacuevas I would like your opinion on this when you return.

[jonnew]

This should really also be done to passthrough devices since these are really local devices that are pretending to be remote.

@aacuevas I believe that in ONI 2.0 we had plans to move the frame clock capture before data arbitration for all devices such that for hub 0 devices, hub clock and frame clock would be identical and for devices that are not on hub 0, frame clock would reflect the time of data capture rather than frame formation. Thus, this may be eventually be obsolete. @bparks13 for this reason, please put a note concerning this in the code. If this occurs then it will have no effect on the actual sample time data for local devices.

[bparks13]

Changing this to a draft until we can understand what is going on better with how the hub clock and frame clock are generated.

[jonnew]

@bparks13 Can you please have somebody at the Allen capture data with two neuropixels 2.0 probes. Ideally this should be done with the bonsai library and all Frame and hub clock data saved so we can examine the jitter. Ideallyt this should be done in saline with a square wave applied via bath electrode, something like 1kHz should be good.

[bparks13]

@jonnew @aacuevas This PR is now ready for review again.

[aacuevas]

Seems pretty straightforward