Add millis functions and serde for unix_timestamp

[kujeger]

Hi!

First off, thanks a lot for this excellent crate.

I've been happily using time for a few projects, but often have the need to deserialize and parse data featuring unix timestamps in millisecond format -- usually created in some java or kotlin application where this seems to be very common.

Instead of only doing local altering/parsing of this, it seems like it could be a good fit along with the existning nanosecond support.

I've tried to make the new code as close as possible to the existing nanosecond/second functions where possible.

[codecov[bot]]

Codecov Report

Merging #529 (cfb94d0) into main (02aa2b8) will decrease coverage by 0.1%. The diff coverage is 74.0%.

@@           Coverage Diff           @@
##            main    #529     +/-   ##
=======================================
- Coverage   98.3%   98.2%   -0.1%     
=======================================
  Files         76      77      +1     
  Lines       8275    8325     +50     
=======================================
+ Hits        8138    8175     +37     
- Misses       137     150     +13     

Impacted Files	Coverage Δ
time/src/serde/mod.rs	96.5% <ø> (ø)
time/src/serde/timestamp_millis.rs	43.5% <43.5%> (ø)
time/src/date_time.rs	94.1% <100.0%> (+0.2%)	:arrow_up:
time/src/offset_date_time.rs	100.0% <100.0%> (ø)

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jhpratt]

Ultimately I want to add a [unix_timestamp] component, which could have support for modifiers to determine if it's second, millisecond, microsecond, or nanosecond. As such I would prefer not to add the serde helpers here.

As to the non-serde methods, there is no reason someone cannot simply multiple the value by 1,000,000 (for milliseconds) or 1,000 (for microseconds) and then pass the value to from_unix_timestamp_nanos. For getting the value, it can be divided by the same amount. For this reason I do not see the need to add these methods at the present time.