compiler: add -O0/-O2 flags, -v 1..4, whole-word JIT, lazy init, auto-inline control

- Add -O0 (no opts) and -O2 (no opts + no checks) CLI flags
- Add -v 1..4 verbosity: timing, per-function stats, debug dump, optimization diffs
- Whole-word JIT: compile Definition bodies to native x86-64 for CT VM
- Lazy CTMemory/JIT/ctype init (defer ctypes until first CT execution)
- Stack-effect comment parsing for stack-checks
- Peephole rules refactored to module-level dicts with O(1) lookup
- _emitted_start short-circuit (O(1) vs O(n) scan)
- Add --no-auto-inline to disable auto-inlining of small asm bodies
- Fix --ct-run-main to execute after linking (prevent SIGSEGV)
- Stdlib: small optimizations like swap + drop -> nip

stdlib/mem.sl

@@ -9,7 +9,7 @@ word alloc
     -1     # fd
     0      # offset
     mmap
-    swap drop
+    nip
 end
 
 #free [*, addr | size] -> [*]
@@ -38,8 +38,7 @@ word memcpy
         c!
         swap
     end
-    swap
-    nip
+    drop
     r> dup -rot - swap
 end