Incorrect documentation for ecdsa::RecoveryId

[zigtur]

At recovery.rs#L45, the documentation indicates:

/// - hi bit (3/4): did the affine x-coordinate of 𝑘×𝑮 overflow the order of /// the scalar field, requiring a reduction when computing r?

However, having hi bit (3/4) is not possible. It should be hi bit (2/3) instead.

[tarcieri]

Yes, feel free to submit a PR to fix it

[geooner]

As this issue has been open for a while now, I've created #970 to fix it. @tarcieri please have a look when you're free.