[LTLToCore] Lower LTL delay/clock patterns to core logic

[robert-at-pretension-io]

Summary
Lowers ltl.delay, ltl.implication, and ltl.clock patterns into core comb/seq logic during the LowerLTLToCore pass. This allows bounded
model checking pipelines (like circt-bmc) to natively verify temporal hardware properties.

Details
Extends the existing LowerLTLToCore pass to translate high-level LTL timing properties into physical hardware monitors (shift registers and
combinational gates). The lowering is implemented safely within MLIR's OpConversionPattern infrastructure (AssertLikeConversion), ensuring
seamless integration with the pass driver's legality checks.

Key Features:

• Exact Delay (##N): Lowers strict delays by shifting the antecedent through N seq.compreg flip-flops and checking implication against the
  consequent.
• Bounded Time Windows (##[N:M]): Lowers bounded windows by generating a "pending request" token monitor. The monitor shifts forward each cycle
  and is dynamically cleared if the consequent is observed while the window is open. A violation is triggered if the token reaches the deadline
  unsatisfied.
• Unbounded Delay: Explicitly unsupported; gracefully returns a clear MLIR diagnostic error (unsupported LTL delay pattern for LowerLTLToCore).
• Dialect Registration: Registers circt::ltl::LTLDialect in circt-bmc.cpp so the tool can safely parse and traverse LTL properties before
  they are lowered to core.

Tests Added

• test/Conversion/LTLToCore/ltl-delay-ok.mlir: Validates the correct generation of shift registers and boolean logic for both exact and bounded
  delays (via FileCheck).
• test/Conversion/LTLToCore/ltl-delay-error.mlir: Ensures unbounded delays are properly rejected and flagged (via --verify-diagnostics).

[TaoBi22]

Hi @robert-at-pretension-io, thanks for taking a stab at this too! It would be great to support these cases, but it looks to me like what you've added to LTLToBMC would probably make sense to have in LTLToCore instead, since they all seem to tackle the issue by modelling the LTL using core dialects (LTLToCore isn't in the circt-bmc pipeline yet, but I've been meaning to add it so I'll open a PR for that to not stall you).

[TaoBi22]

Thanks for taking another pass @robert-at-pretension-io - just some high level comments for now, I'll do another review once they're addressed to avoid any redundant feedback

[TaoBi22]

Thanks for swapping this to use pattern matching infrastructure @robert-at-pretension-io! It looks like now the rewrites are defined as patterns over AssertLike operations, but are actually rewriting ops further up the DAG rather than the AssertLike operations. Could this instead be some patterns on those actual ops? (I think it violates MLIR pattern rewriter restrictions to not actually modify the op the pattern is operating on)