Alternative fes_to_bytes_zeropad with no dropping of last bits

[optout21]

Fixes #198 , as discussed there

[apoelstra]

In 15f6d0e:

The change to the doccomment starting with "If the total number of bits" is wrong. (Though I agree this sentence is really long, and if you can break it up somehow, you should.)

Otherwise ack.

CI failure is unrelated and is fixed by #200 (which I will merge shortly, so you will just need to rebase).

[optout21]

The change to the doccomment starting with "If the total number of bits" is wrong. (Though I agree this sentence is really long, and if you can break it up somehow, you should.)

I changed the documentation, made the explanation more detailed.
Let me know if it's OK; should I squash.

    /// Similar to `fes_to_bytes`, but in this variant trailing bits are kept.
    ///
    /// If the last bits do not fill up the last output byte completely, they
    /// are not dropped, but kept, and the last byte is padded with zero bits.
    /// In effect this is the same as if the input was padded with zero values
    /// (1 or 2), such that there are enough zero bits to fill the last byte.
    /// The output is either the same as of `fes_to_bytes`, or has 1 extra byte.

[apoelstra]

New text looks great! Nah, I wouldn't bother squashing.

[apoelstra]

ACK 06762e0 successfully ran local tests

[apoelstra]

Will let @clarkmoody review before merging since this is an API change.

[clarkmoody]

ACK 06762e0