QsfpModuleController: SCL stretch detection is too naive for our QSFP control use case

[Aaron-Hartwig]

The I2C SCL stretch detection is very simple (

quartz/hdl/ip/bsv/I2C/I2CBitController.bsv

Lines 155 to 161 in e5f3844

	// After the delay we know SCL is being stretched if we aren't the ones
	// holding it low.
	(* fire_when_enabled *)
	rule do_sample_scl_stretch(scl_stretch_sample_strobe && scl_in == 0);
	scl_stretching <= scl_out_en == 0;
	scl_stretch_sample_delay <= False;
	endrule

): When the I2C controller is not driving SCL, wait a small delay and see if someone else is holding it low. The problem in the context of our QSFP controller is that each module's power supply is only enabled when a module is installed. That rail also energizes the pull-up resistors on the module's I2C rail, so the SCL line will go low. We erroneously detect this as a SCL stretch.