fix: zero sequence check for from_height

[ChrisCho-H]

This is following #740.
I missed from_height also needs to check if zero.
I'm afraid as it's API breaking change.

[apoelstra]

CI failure is unrelated.

Concept ACK. I'm tempted to also suggest we provide infallible conversions to/from NonZeroU16, but the API for that type is super annoying, and there's also no equivalent type that can express the allowable range of AbsLockTimes so it'd make our API asymmetric.

@ChrisCho-H can you add a from_height_unchecked method which has the original behavior? You don't need to use it anywhere (it's longer to type than from_height(x).unwrap() after all) but I think some users want to grep for unwrap and have the output be clean.

[ChrisCho-H]

I think NonZeroU16 could be better option for the explicit and infallible usage, but as you mention it would be very inconvenient to use. The only failing case is when zero is given(which is unlikely to happen for the most cases), so we might better not to tighten interface that much.

[apoelstra]

ACK d030b8f successfully ran local tests