ESMValCore: pin on iris<3.0.4 to still allow (formal) support for Python=3.6

[valeriupredoi]

OK here's a twisted one: @zklaus and myself have pinned iris<3.0.4 (the latest release from last week; here's the PR ESMValGroup/ESMValCore#1252 ) because they dropped support for Python=3.6 in that latest version and we would still (formally) support it until we release 2.4. Hence we have released ESMValCore 2.3.1 with that pin in. Here is the associated PR to drop support for Python 3.6 in ESMValCore ESMValGroup/ESMValCore#1200

This pin means we are using a slightly outdated set of dependencies:

• cftime<1.5
• iris<3.0.4
• scipy<1.6

We are actually using cf-units>3.0.0 so this means in practice it is impossible to allow for a Python=3.6 environment solving, so we've already dropped support for it in reality, but we've not yet formalized the drop. Once we do that, we will be able to update all those dependencies, and have a good and modern environment. There are a few things to be done for dropping support for Python=3.6:

• change the documentation
• change tests on Circle and (a few) of Github Actions
• email users

[valeriupredoi]

@bouweandela we have already eliminated support for Python 3.6 right? Do we need to do more on this?

[bouweandela]

It might be nice to unpin iris again so we can use the latest features?

I'm not sure there was ever a need to pin iris, because if you want to use our tool with Python 3.6 you can do so with the older version of iris and if you want to use it with a newer version of Python there is no need to use an old iris version.

[valeriupredoi]

yeah no, was just about to do it today (then I got busy with bureaucracy shtuff at work), it's high time we did it, I'll do a PR now, let the GA tests go nuts too 👍