Remove Aether, go all in on coursier

[johnynek]

I think we can reconceptualize this tool as a fixed lock file resolver for coursier. So we resolve all the dependencies then emit the hashes and the dependency relationships between them to a file and that file (not the jars) can be check in.

One output format would be the current output format, but we could also emit lock files for use with other build tools (e.g. sbt we could make a plugin that uses the lock file and only fetches the files).

What do you think @ianoc

[johnynek]

actually, the pro move would be to solve this:

coursier/coursier#1487

Then make bazel-deps a thin layer that converts such a lock file into bazel consumable data.

[ianoc]

Oh that would be pretty neat, having Bazel dips just possibly be the input yaml format -> outputs needed for bazel and moving the main logic upstream would I think be 💯 . The Aether resolver is buggy enough I can't imagine anyone could be recommended to use it, so seems like little win to keep it around and not simplify/stream line around just courier