Most implementations using 'B' header encoding (e.g. for
ISO-2022-JP) encode 'minimal' portions of header value and
do not encode US-ASCII portions as much as possible.
This patch does this.


Current version (2.4.3) of email.Header module encodes
whole of header value.  Though this feature is 
MIME-compliant, especially with 'B' encoding, causes
some difficulty.

For example: Subject prefix of mailing list postings
are usually
written in ASCII characters -- GL part of its charset.
 Encoded
form of subject prefix is not only human-readable but also 
some search, header-rewrite etc. capabilities are forced to
implement preprocessing for decoding.