Implement MapSkipError, analogous to VecSkipError, but for map-like data structures.

[johnmave126]

Currently, there is a VecSkipError for resiliently handling heterogenous sequences. However, there are also heterogenous maps in the wild, notably the bundle version information plist from Apple.

This PR implements MapSkipError, analogous to VecSkipError, deserializes an entry only if both the key and value are deserializable.

Notable Changes

• The impl of VecSkipError is moved to a new file de/skip_error.rs and ser/skip_error.rs to reuse GoodOrError with MapSkipError.
• The impl of MapSkipError works for all the maps enumerated by impls::macros::foreach_map.

Unanswered Questions

• Do we also need to implement this for kv-tuple lists?
• Do we also need to implement modules under serde_with::rust for BTreeMap and HashMap?

[codecov]

Codecov Report

Attention: Patch coverage is 79.41176% with 7 lines in your changes missing coverage. Please review.

Project coverage is 67.32%. Comparing base (dee706a) to head (94ff4c1).
Report is 7 commits behind head on master.

Files	Patch %	Lines
serde_with/src/de/skip_error.rs	76.66%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #763      +/-   ##
==========================================
+ Coverage   67.25%   67.32%   +0.07%     
==========================================
  Files          38       40       +2     
  Lines        2452     2464      +12     
==========================================
+ Hits         1649     1659      +10     
- Misses        803      805       +2     

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

[jonasbb]

Thanks for the PR, this looks all really good.

Regarding your questions:

1. It is not necessary to support kv-list. You can add support if you want, but this PR is fine to merge as is.
2. No, you shouldn't add more to serde_with::rust. The VecSkipError doesn't have any other variations either. In general, it is difficult to combine such different behaviors and if you combine too many you have a giant compatibility matrix.

[johnmave126]

Cooool.

You are right. Under the current way of enumerating different types of map, adding support to kv-list probably means adding support for MapPreventDuplicates and alike, which is out of the scope of this PR. I don't plan to further change this PR then.

[jonasbb]

Thanks for the work :) I try to prepare a new release this weekend.