fix logic for growing mailbox memory

[jpalus]

• What does this PR do?

Plugs memory leak when growing memory space for mailbox.

• What are the relevant issue numbers?

Fixes #4087

[jpalus]

What I don't particularly like about mx_alloc_memory is that despite its name req_size does not represent size but rather a maximum index to be used in m->emails so size-1. Possibly to fixed/adjusted but better to do it independently.

[flatcap]

What I don't particularly like about mx_alloc_memory...

hehe, there are quite a few areas like this.
Code that's clearly evolved a long way from its original purpose.

Possibly to fixed/adjusted but better to do it independently.

Please. Suggestions welcome.

One of the long-term goals is to finally separate the GUI from the backends.
When that's done, it should be possible to kill mx_alloc_memory().

[flatcap]

I've folded this PR into #4090.
Thanks