Use overflowing_naive_local in DateTime::checked_add*

[pitdicker]

The final part of #1048.

Using overflowing_naive_local in DateTime::checked_(add|sub)_days and DateTime::checked_(add|sub)_months prevents panics.

I thought it would be good if users can rely on another property: that a clear no-op works as a no-op. I.e.DateTime::checked_add_days(Days::new(0)) should always return the original value.
To do that, checked_(add|sub)_days and checked_(add|sub)_months have a fast path that returns Some(self) if it is a no-op.
They already had a fast path anyway, we just start relying on it.

with_year gains a fast path to have the same property.

What we can't promise is that you can add or subtract and create a value that is just out of range for the local time but still in range for the UTC value. That would require alternative methods on NaiveDate for add_days, checked_*_months and with_year that have a different bound checks. It just doesn't seem worth it to me.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.14%. Comparing base (02c68d6) to head (f23dcba).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1333      +/-   ##
==========================================
+ Coverage   92.11%   92.14%   +0.03%     
==========================================
  Files          40       40              
  Lines       18026    18089      +63     
==========================================
+ Hits        16604    16668      +64     
+ Misses       1422     1421       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pitdicker]

@djc May I ask for this PR as a next one to review?
If @Zomtir wants to work on NaiveDate::checked_add* and NaiveDateTime::checked_add* (and with_year...) it would be nice to have these fixes in.

[pitdicker]

Thank you!