Support narrow p-code stack pointers

[zardus]

THIS MESSAGE WAS GENERATED BY AN AUTOMATED PROCESS

ArchPcode derived the initial stack pointer with an expression that assumed a stack pointer at least 16 bits wide, so constructing an architecture with a narrower one — the 8051 and the 8048 have a one-byte stack pointer on a 16-bit address space — raised ValueError: negative shift count.

Deriving the value from the stack pointer's own width instead fixes it where the register sizes are read, and leaves every architecture with a wider stack pointer exactly where it was.

The regression constructs ArchPcode("8051:BE:16:default") and pins the unchanged value for z80, 68000 and x86-64.

Validation: #363 (comment)

Related: angr/cle#717, angr/angr#6793.

[zardus]

THIS MESSAGE WAS GENERATED BY AN AUTOMATED PROCESS

Validation record for head 08d2caf1ef8cf15be8a5fe494dab33a9e45ddf42 against baseline da171ca0dd8eec1e16dd9c4f04576b2c958e2aea.

• Reproducer: python -c "import archinfo; archinfo.ArchPcode('8051:BE:16:default')" — ValueError: negative shift count on the baseline
• Regression: python -m pytest tests/test_pcode.py::TestArchPcode::test_arch_with_narrow_stack_pointer — fails on the baseline with that error, passes on the head
• Full suite: python -m pytest tests — 18 passed
• Lint/type: pylint 4.0.6 and pyright 1.1.411, scored per changed file against the merge base the way the hosted Lint and Typecheck jobs score them — archinfo/arch_pcode.py 10.00 -> 10.00 and badness 0.238 -> 0.235, tests/test_pcode.py 10.00 -> 10.00 and badness 0.0 -> 0.0
• Hooks: pre-commit run --all-files — every hook passes and leaves the tree unchanged
• Coverage: on the head, constructing ArchPcode for every language pypcode 4.0.1 exposes succeeds, with no exception anywhere in the list
• Environment: Python 3.12.13, pytest 9.1.1, pypcode 4.0.1

Caveats: initial_sp is still taken from the address space width when the stack pointer register is not literally named sp (PIC-16 calls it stkptr), and AARCH64:BE:32:ilp32 still gets a value from its 64-bit sp that sits outside the 32-bit address space; both predate this change and are untouched by it.

[angr-bot]

Corpus decompilation diffs can be found at angr/dec-snapshots@master...angr/archinfo_363