Cell comparison: remove ancient netcdftime/cftime workarounds

[rcomer]

🚀 Pull Request

Description

Closes #4697. The example given there now produces

[Cell(point=cftime.datetime(2022, 3, 8, 12, 0, 0, 0, calendar='standard', has_year_zero=False), bound=None), Cell(point=cftime.datetime(2022, 3, 9, 12, 0, 0, 0, calendar='standard', has_year_zero=False), bound=None)]

The first workaround was added at #1016. From the comments on the tests I removed, the comparisons were then falling back to comparing object ids, so presumably rich comparison was not implemented in that version of netcdftime at all. It is implemented now:
https://github.com/Unidata/cftime/blob/dc75368cd02bbcd1352dbecfef10404a58683f94/src/cftime/_cftime.pyx#L1314

The second workaround was added at #2440. The comment there says that NotImplemented was not always being returned by the comparison method. If I've understood correctly, @bjlittle put in a fix for this at Unidata/cftime#81 where, as far as I can tell, python2 was the main complication. @bjlittle's fix was merged in 2018 and we are now pinned to cftime >= 1.5 which is only a year old. So I think we should be safe.

I'm currently targetting main, but I wonder if this should actually go in a patch release? #4697 suggests that this has now become a problem because time cell points have changed from being datetime.datetime to cftime.datetime. This was a breaking change in cf-units v3.0.

---

Consult Iris pull request check list

[trexfeathers]

@pdearnshaw we're not planning a bugfix release for this, so it would instead be released as part of Iris v3.3. Is that OK with you or is this a particularly urgent issue?

[trexfeathers]

@rcomer What's New and I'll merge 👍

[rcomer]

Thanks @trexfeathers!