lib-mail, lib-imap: Optimize parsing large number of address headers

lib-mail, lib-imap: Optimize parsing large number of address headers

Every header was appended to a linked list by walking through the whole
list, causing excessive CPU usage when the list became large enough.
Fixed by changing struct message_part_envelope to use struct
message_address_list, which stores also linked list tail pointers. This
allows quickly appending to the end of the linked list.