Implementing it today would have the same issues: SQLite runs on more platforms than even LuaJIT; and even then I still don't see that a whole general purpose language, no matter how small or embeddable (mostly thinking of Lua here) would compete against SQLite's "be good at SQL-y computation" VM. The main argument for Lua is having humans write it; but the human interface to SQLite is supposed to SQL. I'm not sure I understand the benefit of compiling the SQL down to Lua so it can then be interpreted by the standard Lua implementation...
