Implement checks to verify if we exceeded the max constant for a given index size.

[NotJustAnna]

Throughout the bytecode, integers of multiple sizes (12 bits, 16 bits and 24 bits) are used to index values on constant pools. However, the compiler does not care if a given constant is over the index's size in bits, and invalid bytecode will be generated in this case.

Notable cases:

• InvokeLocalInsn/InvokeMemberInsn uses a 16-bits integer #nameConst to index the string pool, whereas the rest of the code uses 24-bits integers.

• PushExceptionHandlingInsn/PushLoopHandlingInsn uses a pair of 12-bits integer (don't judge, size constraints) to index two labels, whereas JumpInsn uses a 24-bit integer.

• Implement checks to verify if a given compilation exceeds either the absolute maximum constant size.

  • In reality, this only matters for the case when we do NOT want to immediately execute the compiled code, so generating the error at serialization time should be okay.

• Reduce the integer used in constant pools/function definitions size to 24-bit integers.