Consolidate CI

[nyonson]

Folded the readme docs inline since those have a tendency to deviate despite our best efforts.

[tcharding]

docs and docsrs I don't mind but I'm not sold on squashing fmt and lint. Same question more or less as #64 (comment)

[nyonson]

docs and docsrs I don't mind but I'm not sold on squashing fmt and lint. Same question more or less as #64 (comment)

Yea, I might be over complicating things here. For a lot of these non-test commands, the overhead of loading up a toolchain is the majority of the "ci minutes" of the job. So was thinking for efficiency it makes sense to amortize that initialization cost for some of the tasks. But if we don't care about ci minutes or # of jobs, maybe just following the "one thing per job" rule is simple enough while also being the easiest to debug failures?

[tcharding]

So was thinking for efficiency it makes sense to amortize that initialization cost for some of the tasks. But if we don't care about ci minutes

We don't, just thrash the CI machines.

or # of jobs,

In rust-bitcoin we tried to stay under 20 because that is all we get in parallel on the free tier and I personally don't like waiting for CI. But now its slow as hell still so in my head I have just given up on this goal.

maybe just following the "one thing per job" rule is simple enough while also being the easiest to debug failures?

Ease of debugging is the most important. Dev time is expensive, machine time is cheap.

[tcharding]

ACK f783c1d