dovecot-pigeonhole.m4: Make testsuite available when linked against Pigeonhole source tree, even when it is not yet compiled.

imapsieve plugin: During transaction run the synchronized version of the target mailbox was opened too early.

This meant opening a mailbox a second time even if it did not match any mailbox rules.

lib-sieve: edit-mail: Passed FALSE for normalizer argument of message_header_decode_utf8() rather than NULL.

imapsieve plugin: Restructured imap_sieve_run_scripts() function such that the compile error code is never used uninitialized.

lib-sieve: variables extension: Changed declaration of new variables to avoid useless memory allocation.

Also prevents assert failures with new hash_table_insert() assertion against existing entries.
This changes the API for the variables extension.

This prevents trouble with o_stream_nsend_istream() later.

lib-sieve: Changed sieve_extension_override() to use hash_table_update() rather than hash_table_insert().

Due to a Dovecot change this would now cause an assertion failure.

managesieve-login: proxy: Did not handle zero return value from managesieve_parser_read_args() correctly.

This would uselessly allocate a new entry with the same content.

lib-sieve: variables extensions: Fixed handling of importing a variable into a scope that is already imported earlier.

This allocated a new variable entry every time.

Plain t_malloc was deprecated in dovecot.

This could have happened with Maildir/mbox and autoexpunging.
Mimics Dovecot/IMAP change ddffbb277e212adb9d594b3d7d1e1d39fa78f7c6.

Mimics Dovecot/IMAP change 1001bda65e9f592d852a26ddd6a993f7973050bf.

The recipient_delimiter is treated as multiple one-character delimiters
rather than one multi-character delimiter if more than one character is
supplied.

Based on a patch by: Lennart Weller <lhw@ring0.de>

Stream errors were never logged properly.
Streams now have a name that makes sense.
Timeout errors were handled in a rather stupid fashion.

The previous change prompted some futher code review and similar improvements/fixes were applied at various places in the code.

If header was truncated, it was treated as an istream error. Also the
istream read errors weren't logged.