Simplify SerdeError

[pitdicker]

While investigating all the CI failures on #1454 I noticed our serde code tries to convert LocalResults to a custom error type.
This can all be done in a much simpler way.

I did keep around a simple error enum because it allows reporting the invalid value.
It currently has only one variant for deserialization errors for UNIX timestamps. We could simplify it further, or leave room for adding more formats to deserialize.

Also included is the fix from #1396 for an invalid cast in the ts_seconds module for NaiveDateTime.

[codecov]

Codecov Report

Attention: Patch coverage is 65.85366% with 14 lines in your changes are missing coverage. Please review.

Project coverage is 92.12%. Comparing base (02c68d6) to head (bc03a0f).
Report is 3 commits behind head on main.

Files	Patch %	Lines
src/lib.rs	0.00%	10 Missing ⚠️
src/datetime/serde.rs	90.47%	2 Missing ⚠️
src/naive/datetime/serde.rs	80.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1458      +/-   ##
==========================================
+ Coverage   92.11%   92.12%   +0.01%     
==========================================
  Files          40       40              
  Lines       18026    18000      -26     
==========================================
- Hits        16604    16583      -21     
+ Misses       1422     1417       -5     

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

[djc]

Nice!

[djc]

Three suggestions:

• Maybe this function should be attached to SerdeError as a sort of (not quite) constructor, like SerdeError::invalid()?
• (If not, this function should be defined above SerdeError)
• Let's move this into the crate::serde inline module in lib.rs?

[pitdicker]

• Maybe this function should be attached to SerdeError as a sort of (not quite) constructor, like SerdeError::invalid()?

• (If not, this function should be defined above SerdeError)

In my defence, that function looked very different when I started 😆. But I'll keep trying to put things in that order.

I like that with a standalone function only invalid_ts is public (within the crate), with the type just an implementation detail.

• Let's move this into the crate::serde inline module in lib.rs?

Good suggestion, much more logical place. I added the move as an initial commit.