kens are collapsed into single spaces in values: in
the value of a 'bare-quoted-string' (a quoted-string with no leading or
trailing whitespace), any whitespace that appeared between the quotation marks
is preserved in the returned value.  Note that in all Terminal strings quoted
pairs are turned into their unquoted values.

All TokenList and Terminal objects also have a string value, which attempts to
be a "canonical" representation of the RFC-compliant form of the substring that
produced the parsed subtree, including minimal use of quoted pair quoting.
Whitespace runs are not collapsed.

Comment tokens also have a 'content' attribute providing the string found
between the parens (including any nested comments) with whitespace preserved.

All TokenList and Terminal objects have a 'defects' attribute which is a
possibly empty list all of the defects found while creating the token.  Defects
may appear on any token in the tree, and a composite list of all defects in the
subtree is available through the 'all_defects' attribute of any node.  (For
Terminal notes x.defects == x.all_defects.)

Each object in a parse tree is called a 'token', and each has a 'token_type'
attribute that gives the name from the RFC 5322 grammar that it represents.
Not all RFC 5322 nodes are produced, and there is one non-RFC 5322 node that
may be produced: 'ptext'.  A 'ptext' is a string of printable ascii characters.
It is returned in place of lists of (ctext/quoted-pair) and
(qtext/quoted-pair).

XXX: provide complete list of token types.
é