Limit the precision of floating point event fields

[jyasskin]

There's a history of factory calibration information being used to fingerprint individual devices. I haven't investigated whether this is possible on common pointing devices, but as a precaution, would it make sense to establish a normative limit on the precision of the fields in pointer events, similar to what was done for deviceorientation and accelerometer? See w3c/deviceorientation#86, which limited angle measurements to 0.1 degrees.

[mustaqahmed]

Looks like we should be fine with a 0.01 precision in all of these double attributes and the inherited coordinates. Or maybe even 0.1 in some cases. Thoughts?

We don't need to "downgrade" the IDL types to float, do we?

[jyasskin]

I think float fields are also vulnerable to this problem, so they should also be limited, and there's no need to consider changing doubles to floats.