Timestamp Serialization Should Not Be Fixed to Nanoseconds

[almann]

#272 ports some internal code to APIs, part of this is a TODO to make Timestamp serialization to be not fixed to nanoseconds.

When searching for prior issues, I found #17, but I am not sure if that is different or already resolved (since it indicates skipping fractional seconds). @zslayton can probably confirm this.

---

So this means that if there exists sub-second precision, that precision is always nine digits (i.e. nanoseconds)?

Originally posted by @almann in #272 (comment)

[zslayton]

Prior to #272, the binary reader/writer only worked with DateTime<_>, which represents its fractional seconds as nanoseconds. Both reading and writing had a TODO to support other levels of precision.

Now that we're supporting reading and writing Timestamps (which natively support setting/getting the precision), assuming nanoseconds on this code path no longer makes sense.