Add memory size syntax to HLIL load/store operations

Version and Platform (required):

• Binary Ninja Version: 5.1.8104 Personal (6eaaf348)
• Edition: Non-Commercial
• OS: Windows
• OS Version: Windows 11 24H2 Build 26120.5770
• CPU Architecture: x64

Bug Description:

The sh instruction in the following assembly is incorrectly decompiled as *(i_0 - 2) = ..., even when the type of i_0 is already uint16_t *.

nop     
lhu     $a0, 0($v0)
daddiu  $v0, $v0, 2
xor     $a0, $a0, $a1
sh      $a0, -2($v0)
bne     $v0, $v1, 0x234
nop     

Steps To Reproduce:

1. Open the attached BNDB.
2. Go to address 0x00000244.
3. Compare the HLIL and disassembly.

Expected Behavior:

The decompiled HLIL should be like *(i_0 - 1) = ... or i_0[-1] = ....

Screenshots/Video Recording:

Binary:

Binary Ninja Portal file uploader ref words:

divine axis tokenizes vividly

Additional Information:

N/A.

Start with the attached BNDB and inspect the HLIL and disassembly at address 0x00000244, focusing on the load/store operations for the shown MIPS instructions. Compare the current pointer offset with the expected *(i_0 - 1) or i_0[-1] form; done means the HLIL reflects the element-sized memory offset.