[Experimental][ResourceSharing] Introduce ResourceSharingFCCM22Pass

[dapfister313]

This PR introduces the ResourceSharingFCCM22Pass. This implements the methodology described in paper "Resource Sharing of Dataflow Circuits" presented at FCCM '22. There is the implementation of the complete algorithm that finds a grouping scheme for operations of the same type, such that there is no performance degradation when a group of operations share the same physical functional unit.

For example, the following outputs the best possible grouping scheme for the multipliers:

dynamatic-opt handshake_transformed.mlir \
    --handshake-set-buffering-properties="version=fpga20" \
        --sharing-FCCM22="algorithm=fpga20 frequencies=$F_FREQUENCIES timing-models=$DYNAMATIC_DIR/data/components.json timeout=300 dump-logs" \
    > "handshake_buffered.mlir"

[lucas-rami]

Thanks for the contribution! I haven't had the time to look at it beyond a very far away glance (and will most likely not have time before the beginning of March unfortunately, having important deadlines on my side), but I already have a couple comments.

• This is a huge contribution (2000 LoC), which makes it more work for the both of us.
  • This amount of code takes me a couple continuous hours to review thoroughly (and it only really makes sense for me to review things in one go) instead of multiple and more manageable chunks of time. That's a harder to schedule piece of work for me.
  • If I find something in the design that is really sub-optimal and has to be changed before the PR can be approved, this change now potentially has to reverberate in 2000 LoC instead of, say, 500. That's more work for you.
• I don't know if the code is clang-formatted, but it is at least not clang-tidied (static code analyzer) because I see variable/function names following the snake_case syntax when the project is configured to report anything that is not camelCase (except in specific contexts). Please install these tools on whichever IDE you are using, run them, and fix all warnings that clang-tidy generates. On VSCode both clang-format and clang-tidy are provided by the llvm-vs-code-extensions.vscode-clangd extension.
• The PR has conflicts with main as GitHub indicates, this needs to be fixed.

For reference, all of these recommendations are in our contribution guidelines.

[dapfister313]

I'm not entirely sure, but I think, I resolved the conflicts.