Consider format-specific timestamp decode methods

[SeanTAllen]

While planning the implementation of #13 (format-specific bin/ext/str decoding), we noticed a similar gap in the timestamp family.

The encoder has format-specific timestamp_32, timestamp_64, and timestamp_96 methods, but the decoder only has the compact timestamp() method, which uses an if chain to dispatch across the three wire formats (fixext_4, fixext_8, ext_8). The same performance argument from #13 applies here — callers who know the exact timestamp wire format pay for unnecessary dispatch.

That said, timestamps are a higher-level concept layered on top of ext wire formats (they use ext format bytes with ext_type 0xFF), so they're arguably a different category than the "raw format-specific decode" methods in #13. Some questions worth considering:

• Should format-specific timestamp decode methods (timestamp_32, timestamp_64, timestamp_96) be added for symmetry with the encoder?
• Is the performance benefit meaningful for timestamps specifically, or is the if chain negligible given the data sizes involved?
• Should this be part of Add "faster" bin/ext/str decoding #13's implementation or a follow-up?

Kicked out from the #13 planning process to consider separately.