fix: support arbitrary types with custom __eq__ and Annotated with python 3.9

[PrettyWood]

Change Summary

When the default value is an instance of a class with a custom __eq__, it can fail as shown in the written test.
But the question is "why did we switch from an identity check against Undefined to an equality check against Undefined AND Ellipsis?`
The answer lies here
https://github.com/samuelcolvin/pydantic/blob/62bb2ad4921016df51abf3922c3fe51113b08939/tests/test_annotated.py#L121-L130

To support this, a check on Ellipsis has been added, which is wrong IMO.
It has been done because the default value of the Field is not the same once it has been validated (Undefined becomes Required aka Ellipsis)

2 solutions:

• store also the default value of the field in original_default and use this one for the check against Undefined
• do not change the default value and keep Undefined

I went with solution 2 as it doesn't seem to have any impact

Related issue number

closes #2483
closes #2639

Checklist

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.md file added describing change
  (see changes/README.md for details)

[codecov]

Codecov Report

Merging #2502 (bdd0ea8) into master (7b7e705) will not change coverage.
The diff coverage is 100.00%.

❗ Current head bdd0ea8 differs from pull request most recent head 683b0c1. Consider uploading reports for the commit 683b0c1 to get more accurate results

@@            Coverage Diff            @@
##            master     #2502   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           25        25           
  Lines         5109      5108    -1     
  Branches      1050      1050           
=========================================
- Hits          5109      5108    -1     

Impacted Files	Coverage Δ
pydantic/decorator.py	100.00% <100.00%> (ø)
pydantic/fields.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 14f055e...683b0c1. Read the comment docs.

[samuelcolvin]

awesome, thank you so much.