Final fix for emplace

[monarchdodra]

This should be the final fix for emplace.

It fixes support for static arrays, as well as for structs with const members. It also consolidates the code somewhat, and has even more unittests.

This should fix the new issues 11173 and 11178
http://d.puremagic.com/issues/show_bug.cgi?id=10973
http://d.puremagic.com/issues/show_bug.cgi?id=10978

[monarchdodra]

@klickverbot : With this last pull, I think emplace should be "mostly completely fixed". You've been reviewing and merging these so far. Could you review this last one?

It also simplifies and relaxes the implementation, so should still work even if/when kenji introduces his newer and "more powerful" constructors.

FYI, it addresses 10973, which is a regression I introduced, and Walter wanted either fixed or reverted for 2.064. I'm opting for the "fix" solution :)

[DmitryOlshansky]

@monarchdodra you may want to link pulls to bugzilla. Usually proves to be very helpful.

[dnadlinger]

@WalterBright: One of the issues was a regression.

[WalterBright]

Pulling a regression fix does not automatically pull it for 2.064. This means that these fixes can get overlooked for 2.064. They need to be cherry-picked into 2.064.

[WalterBright]

I cp'd this for 2.064.

[dnadlinger]

@WalterBright: This is why I pinged you. Or do you mean not closing the bugzilla issue until the relevant course of action for 2.064 has been decided on?