LLVM changes related to BPFv2

[pgarg66]

Problem

This issue is split off from #20323 Need LLVM changes corresponding to the following items

• [ ] Explicit sign extension of results (https://github.com/solana-labs/rbpf/pull/548) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/83
• [x] Replace sub r11, imm with add r11, -imm (https://github.com/solana-labs/rbpf/pull/488) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/70
• [x] Merge sub reg, imm and neg (https://github.com/solana-labs/rbpf/pull/489) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/73
• [x] Replacement for lddw (https://github.com/solana-labs/rbpf/pull/486 and https://github.com/solana-labs/rbpf/pull/496) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/74
• [x] Fix encoding of callx (https://github.com/solana-labs/rbpf/pull/491) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/78
• [x] Disable little-endian byte swap (https://github.com/solana-labs/rbpf/pull/493) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/75
• [x] Wide multiplication, signed division (quotient / remainder) instructions: two 64 bit factors resulting in a 128 bit product (https://github.com/solana-labs/rbpf/pull/498) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/81
• [x] Remove the concept of "input buffers" and its associated instructions (https://github.com/solana-labs/rbpf/pull/251): ldabsw, ldabsh, ldabsb, ldabsdw, ldindw, ldindh, ldindb, ldinddw - LLVM PR: https://github.com/solana-labs/llvm-project/pull/75
• [x] Relocation-less static syscalls (https://github.com/solana-labs/rbpf/pull/296) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/79
• [x] Dynamic stack pointer instead of fixed stack frames (https://github.com/solana-labs/rbpf/pull/274) - LLVM PR: https://github.com/solana-labs/llvm-project/pull/26

Proposed Solution