Clean up and fix tests

[simsurace]

Checked out the registry at the time of the v0.6.5 release and then set the compat upper bounds to the resolved dependencies.

Supersedes #124

[coveralls]

Pull Request Test Coverage Report for Build 7624664125

Warning: This coverage report may be inaccurate.

We've detected an issue with your CI configuration that might affect the accuracy of this pull request's coverage report.
To ensure accuracy in future PRs, please see these guidelines.
A quick fix for this PR: rebase it; your next report should be accurate.

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall first build on freeze-deps at 87.549%

---

Totals
Change from base Build 7303152969:	87.5%
Covered Lines:	1343
Relevant Lines:	1534

---

💛 - Coveralls

[simsurace]

Yay, this works. I will include the other changes I made in #124 to clean things up.

[willtebbutt]

This is fantastic. Would you mind checking to see whether all of the upper bounds are actually necessary? I can definitely imagine that the Zygote / ChainRules ones are, but it seems surprising to be that it's necessary for e.g. FillArrays.

[simsurace]

That's what I was trying to avoid ☺️, I have spent way too much time trying to find such combinations, so that's why I went with the historical registry approach.I think some of the failures are due to JuliaArrays/FillArrays.jl#271. There was a long Slack thread you maybe did not see, I forwarded it to you on Slack.

[willtebbutt]

Ahh okay. No, I didn't see that thread. Could we at least try un-pinning AbstractGPs and KernelFunctions? I would be completely amazed if non-breaking changes in them actually break AD in TemporalGPs.

[simsurace]

Let's try

[simsurace]

This works. Let's then try removing some of the others.

[simsurace]

Ok, now I'm feeling bold. Let me push it a bit further.

[simsurace]

@willtebbutt ok, this is more of a minimal set of pinnings that works.
Did you have any chance to look at the Slack thread? We might as well look at resolving the underlying issues first, but I don't know where to start.

[willtebbutt]

I've had a look at the slack thread, but I'm not going to have a chance to dig into this for a while. I'm happy to approve this PR though

[simsurace]

Thanks. But I think if we merge this right now, perhaps we should put a note at the top of the README, that this package is currently pinning some AD deps and might hold back environments or cause conflicts.

[willtebbutt]

Another idea:

1. release this patch (0.6.7)
2. immediately make another patch release in which we unpin everything (0.6.8)
3. add a note to the readme saying that version 0.6.7 has pinned dependencies, and we're confident that everything should work on that version, so if something if the current version isn't working for you, please try that version out?