* providers/smtp/camel-smtp-transport.c (smtp_data): Remove use of
	linewrap filter.  Headers are now wrapped.  encode_8bit already
	enforces a 998 octet line limit.
	(smtp_data): Also fixed a memleak, we always have to unref our own
	copy of the filters.  We also dont need to remove them manually,
	so dont bother.  The type's an int too ...

	* camel-internet-address.c (internet_unformat): When scanning past
	quotes, remove them also.
	(camel_internet_address_format_address): If the name contains "'s,
	or ','s then strip and quotes and wrap the whole lot in one set of
	quotes.

	* Makefile.am (noinst_HEADERS): We dont want to install
	camel-charset-map-private.h, ever.  There are probably other
	similar files ..?

	* camel-mime-part.c (write_to_stream): Fold header lines
	appropriately as we're writing them out.

	* camel-mime-utils.c (header_fold): Add a new argument, headerlen,
	tells it how long the associated header token is.
	(header_fold): Also,k check to see if we need to fold first, using
	a better algorithm, and also accept already-folded lines, and
	re-process accordingly.
	(rfc2047_decode_word): Add a little buffer space to iconv output
	for shifting overheads?
	(rfc2047_decode_word): finish the iconv with a null call, to flush
	shift state, etc.
	(rfc2047_encode_word): Attempt to break up long words into
	appropriately sized, independent, chunks.  See rfc2047, section 2.
	(header_decode_mailbox): Dont add in extra spaces into the output
	if we are decoding adjacent encoded words.  We can only guess this
	case, as some broken mailers put encoded words inside quoted
	words.
	(header_encode_phrase): Dont merge words if they are going to end
	up too long.  Also change back ot only merge consecutive words of
	the same type.  e.g. 'foo. blah fum.' -> "foo." blah "fum."  or
	'iam an. idiot' -> iam "an." idiot

svn path=/trunk/; revision=6902