Replace upvalue-based shell dispatch with OP_LUSH opcode

Add a dedicated OP_LUSH A B instruction that loads shell C functions
directly from the registry, eliminating the need for upvalue allocation
in mainfunc(), upvalue population in lua_load(), and the fragile
LUSH_NUM_UPVALS heuristic. Every chunk no longer carries 4 extra
upvalues.

lopcodes.c

@@ -102,6 +102,7 @@ LUAI_DDEF const lu_byte luaP_opmodes[NUM_OPCODES] = {
  ,opmode(0, 0, 1, 0, 0, ivABC)		/* OP_SETLIST */
  ,opmode(0, 0, 0, 0, 1, iABx)		/* OP_CLOSURE */
  ,opmode(0, 1, 0, 0, 1, iABC)		/* OP_VARARG */
+ ,opmode(0, 0, 0, 0, 1, iABC)		/* OP_LUSH */
  ,opmode(0, 0, 0, 0, 1, iABC)		/* OP_GETVARG */
  ,opmode(0, 0, 0, 0, 0, iABx)		/* OP_ERRNNIL */
  ,opmode(0, 0, 1, 0, 1, iABC)		/* OP_VARARGPREP */