[HW] Pass to inline modules

Nobody has claimed this yet.

As discussed in the ODM today, it might be generally useful to have a pass for inlining at the HW dialect level. @youngar pointed out that we may already be able to do inlining at the HW level, using the upstream MLIR pass for inlining in the software sense, but that this would inline all modules. FIRRTL has a more fine grained mechanism for inlining. I'm less familiar with that, but I think the gist is there is an annotation we can lay down that marks modules to inline, and a transformation to make it happen. Perhaps what we want here is a similar mechanism, but on the HW dialect.

Start by comparing the upstream MLIR inlining pass with FIRRTL's finer-grained annotation and transformation mechanism. Review the HW dialect's existing module representation and determine how selective module inlining should be expressed. Done means an agreed HW-level mechanism can inline marked modules without inlining every module.

Written by the indexing model from the issue text.