[HW][circt-synth] Implement AggregateToComb pass and add to circt-synth pipeline

[uenoku]

This PR adds a new pass HWAggregateToComb that lowers array operations to comb operations. The pass converts:

• hw.array_get to a mux tree using comb operations
• hw.array_create and hw.array_concat to comb.concat
• hw.aggregate_constant to comb.concat of individual constants

The PR also moves two utility functions from CombToAIG to the Comb dialect:

• extractBits: Extracts individual bits from a value
• constructMuxTree: Builds a multiplexer tree from selectors and leaf nodes

The pass is required to run before CombToAIG to ensure array operations are
properly lowered to AIG. Strictly speaking other than array_get operations can be preserved
but array values prevent optimizations at AIG level. So for the simplicity, I think it's reasonable
to lower array operations within circt-synth pipeline.

I currently added the pass under HW/Transforms considering FlattenIO is very similar but the pass might better belong to Conversion.

[maerhart]

Would it make sense to have aggregate to comb lowering in a separate pass? To me, it seems like it could also be useful outside of the AIG lowering path.

[uenoku]

Would it make sense to have aggregate to comb lowering in a separate pass?

Generally yes, I agree. One concern is that the lowering pattern for HWArrayGetOpConversion might be too low-level for general use case. ArrayGet is lowered into mux-chains for each selector bits but some downstream pass might prefer chains of comb.icmp. Anyway I think having HWAggregateToComb looks great.

[uenoku]

Unnecessary include