Reorganize the JIT optimization passes.

[markshannon]

Currently the JIT works in the following passes:

1. Create a trace (by projecting from the current point of execution)
2. Convert global variables to constants
3. Type based redundancy elimination
4. Peephole redundancy elimination
5. Create executor

What we want is:

1. Create a trace and redundancy elimination
2. Partial evaluation
3. Create executor

To get there we need to combine current steps 1-4 into a single step and implement the partial evaluator. The partial evaluator is discussed elsewhere, this issue is about combining the earlier passes.

To do this, we want to:

• Refactor the symbol structs used in the type analysis to make them more robust and easier to use. See Refactor the JIT's type-lowering/specialization pass's data structures. faster-cpython/ideas#706
• Merge passes 2-4 into a single pass
• Merge that pass into pass 1, using the type information to guide trace projection

Linked PRs

• GH-128939: Refactor JIT optimize structs #128940