Remove jinja2 and pooch from sys_info

[cbrnr]

Fixes #12387.

[larsoner]

Works for me. I think in principle maybe we could want all core dep versions listed somewhere, but in practice over the many years we've had this I don't think these have ever been useful to have reported. So I'm okay with removing them in favor of trimming things down a bit. But @drammock you had suggested possibly moving them rather than removing them so I'll let you look before merge

[cbrnr]

I'm also in favor of trimming down the output to things that will be relevant for us. It is unlikely that a bug is caused by pooch or jinja2 at the end of the day. Of course, we already have a dependencies="developer" option, where we could also move these packages to.

[larsoner]

Of course, we already have a dependencies="developer" option, where we could also move these packages to.

Works for me and might satisfy @drammock's desire to keep them somewhere (he had suggested misc I think but maybe lumping under developer is good enough)

[drammock]

including under developer sounds ideal to me. IIRC there was a brief period when pooch version mattered, which could in theory happen again (e.g., if they decided to do a major API change)

[cbrnr]

I added the other required packages to a new section called "Infrastructure". Well, all but lazy_loader, which doesn't have a version attribute.

[drammock]

pushed a commit to hopefully fix Circle, which seems to have been broken by docdict changes in #12195

cc @withmywoessner; see 43094a7; I think the only necessary changes were the ones relating to blank lines, but I took the opportunity to clean up the formatting a bit too

[drammock]

reported the missing lazy_loader version upstream: scientific-python/lazy-loader#96

[drammock]

pip-pre failure is being tracked in #12413

[larsoner]

Thanks @cbrnr !