Safer publication of the resolved LF

[cl4es]

Structuring so that the LF::resolve returns the membername, which is then properly installed into the outer mh by means of mh::updateForm. This is a plausible or at least partial fix to a safe publication issue where the LF might end up in an inconsistent state, and with this change make test TEST=jdk_foreign has passed 10 times in a row for me. To be ideal we should refactor compileBytecode et.c. to not side-effect lf.vmentry etc directly. We should also re-examine isResolved since it might be redundant with this protocol