we check preamble exists or not more rigorously to avoid to misread
broken(?) mulitpart data.

align types to be lowercase

"broken" implies the folloing type message:
* It says Content-Type: multipart but has no multipart block within it.
* multipart without close-delimiter

our analyzer loop does
do { alloc a new part + delimiter
} while ( not reach the end and non-zero data part ).

modify _get_next_pos() which returns (maxlen, maxlen) if broken
condition is satisfied. It means the next part has zero data, so
the next "alloc a new part" process is ignored in analyzer loop.

fix "text/rfc822-headers" type handling since
duplicated "text/rfc822-headers" parts may appear.
we check the duplication always.

bug fix to parse Content-Type: ... phrase

add several debug codes anywhere.

define rfc822_message_body_head() which uses new find() method.  It
returns the head object of body parts in a chain of Mail::Message
objects.

define rfc822_message_body() is an alias of rfc822_message_body_head().

Content-Type:.*boundary= is case insenstive.

size() checks the given object type. Its check is fundamental but may
be effective.

modify data_type() to search data_type in $self.

define header() and data() methods to retrive $self's header and
data (body) part.

redefine _header_mime_boundary() and _header_data_type() as internal
method and nuke public mime_boundary() method.