fix(guest): take the zero-copy input via the safe get_private_input_slice

[MauroToscano]

Review fix on top of #886, targeting that branch so it lands inside it.

The zero-copy read is the right call — get_private_input() really does copy the
whole input for nothing. But syscalls::get_private_input_slice already borrows
the mapped region in place and returns &'static [u8], and get_private_input
is literally that call plus a to_vec(). So dropping to the slice is the entire
win, without the pointer plumbing:

let input = lambda_vm_syscalls::syscalls::get_private_input_slice();
let input = rkyv::from_bytes::<ProgramInput, Error>(input).unwrap();

Three things that buys beyond the line count:

• No raw pointers in guest code. syscalls.rs keeps the region layout and its
  one unsafe block in a single place — the stated reason
  get_private_input_slice exists. Re-reading the length prefix in the guest
  duplicates layout knowledge that has to stay in step with the executor.
• Restores the length-prefix clamp. get_private_input_slice bounds the prefix
  by MAX_PRIVATE_INPUT_SIZE; ef_io::read_input returns it raw. The executor
  rejects oversized inputs, so honest runs are byte-identical either way — but a
  forged prefix built a slice reaching past the region instead of a bounded one.
  Robustness, not soundness: private input is prover-chosen by definition.
• Drops a dependency on unspecified behavior. ef_io::read_input documents
  buf_ptr as unspecified when buf_size == 0, and the previous code fed it to
  from_raw_parts regardless. Harmless in practice (the implementation always
  writes it, and ethrex input is never empty), but not a contract to lean on.

bench_vs/lambda/recursion already reads its blob exactly this way.

The measured win should be unchanged — both forms skip the same to_vec() — but
it is worth re-running the cycle counts from #886 to confirm, since that number
is the whole point of the PR.

Verified locally

• make executor/program_artifacts/rust/ethrex.elf — builds clean.
• cd tooling/ethrex-tests && cargo test --release -- --include-ignored --skip test_ethrex_real_block
  (CI's exact command) — 4 passed, 0 failed. test_ethrex executes the rebuilt
  guest ELF and compares its public output against a native execution_program
  run, so the changed read path is covered.
• rustfmt --check clean.

Not re-measured: the cycle deltas, and proving-level coverage
(test_prove_ethrex_empty_block is #[ignore] and no CI job selects it).