This rev caused 6 regressions on shell jsreftest only: e4x\decompilation\regress-429249.js ecma\extensions\trapflatclosure.js js1_7\decompilation\regress-429252.js js1_5\extensions\regress-431428.js js1_5\extensions\regress-429264.js js1_5\extensions\regress-422137.js

The recompiler wants to call ReleaseScriptCode to invalidate a script. This used to set script->execPool to NULL, which told us that the script had no compiled code, so we don't need to free that stuff again. With the regressing changeset, script->jit now indicates jit-code is present, but it is not nulled by ReleaseScriptCode. So we go ahead and try to free its member fields, which are garbage. The fix just nulls out script->jit again.