Optimizing Reference Counting with Unique Reference Tracking in JIT

Reference counting is a memory management technique where each object keeps track of how many references point to it. In JIT compilation, optimizing reference counting by identifying uniquely referenced objects allows the compiler to skip unnecessary increments and decrements, reducing runtime overhead. Unique reference tracking detects when an object has exactly one reference, enabling safe elimination of redundant reference count operations.

This PR introduces unique reference tracking specifically for tuples created via the `_BUILD_TUPLE` opcode in the JIT optimizer. By analyzing the JIT's micro-operations, it marks tuples as uniquely referenced unless they are aliased or duplicated through `_LOAD_FAST_*` or `_COPY` operations. Consequently, the optimizer removes redundant reference count increments and decrements for these uniquely referenced tuples during unpacking, such as in `x, y, z = f()`. This targeted optimization reduces the reference counting overhead without compromising correctness.

Understanding and applying unique reference tracking in JIT compilation leads to more efficient generated code by minimizing unnecessary reference count operations, which can significantly improve performance in reference-counted languages like Python. It also lays groundwork for extending these optimizations to other object types beyond tuples.

Review the changes made to handle `_LOAD_FAST_*` and `_COPY` operations in this PR that mark tuples as no longer uniquely referenced. How would you extend this unique reference tracking approach to other container types, such as lists or dictionaries, within the JIT optimizer? Consider what challenges might arise and how you would address them.