I2C: A couple of fixes around SCL stretching is handled

[Aaron-Hartwig]

This commit fixes a couple of bugs related to SCL stretching. First off, in the I2CBitController I adjusted how we sample SCL to detect stretching so we're not errantly sampling. Then in the I2CCore I added a mechanism to handle the SCL stretch timeouts when we see them. Before we lacked this and that meant if that was seen, we were wedged. The rest of the changes were to test cases in order to mimic the behavior observed that brought the problems to our attention.

Fixes #246
Fixes #247

[Aaron-Hartwig]

@nathanaelhuffman I actually did not end up faking out the I2CCore per our conversation. In the event a module would get removed while it was being talked to we would detect that as a timeout given the bus voltage will collapse. However now with an actual timeout recovery mechanism in place we can operate as expected when a module comes back since all error/timeout information only pertains to the last transaction.

[nathanaelhuffman]

lgtm