This reflects the fact that positions stored in matches are
absolute positions, while Position type is used in the
language as position relative to current position in the
sentence context. So mixing the two was confusing at the very least.
This also actually fits the most recent version of spec,
which no longer mentions Position, either.

It was looking for minimum value and Position::Nowhere is
the smallest you can get value-wise (min_int). So if there was
an empty Match in a vector, first_token for the vector would
always end up returning Position::Nowhere, which was wrong
(empty submatches should've been ignored and Position::Nowhere
returned only if there were no non-empty alternatives).

Also, adding M in grammar to be shorthand for $m:_M (vargetter) op in match context.

MatchVector, AnnotationMatch, TokenMatch are now MatchData
MatchData is value_type of Match which now has get_value/set_value
Match has also copy construction working properly now
Default value for Match holds MatchVector so adding $m:_M var in grammar works as intended.

First draft of Match value type with some tests. MatchVector does not yet properly implement first/last token, hence a test fails, and AnnotationMatch is pretty much a stub.

It is coherent now with the SentenceContext::is_inside(int abs_pos)
- SentenceContext::is_inside(Position pos) is added, while
Position::is_inside(SentenceContext) is removed

Values cleanup and doxy: remove some redundancies, make the Begin/End constants in Position an enum, change tag_ref to contents

Add the tagset-symbol-getter operator (GetSymbols), bump required Corpus2 version to 1.0.2 since new features are needed.

It is more intuitive that way, e.g. equal(2,3) is never true, equal(3,3) is always true.
Nowhere is only equal to Nowhere, and Begin or End are equal to nonspecial positions
depending on current position in the sentence context.

* reset_values function to set all Values to their defaul-construct state
* read-only access to all variables of some type via get_all<T>
* Variables cloning
* convenience functions: get_or_throw, get_value
Value subtypes now contain a value_type typedef and require a const value_type& get_value() const member function.

My relative position tests wouldn't compile in GCC where I used Position::Nowhere for comparison. Apparently according to standard TC1, section 9.4.2, paragraph 4, if you want to use integral static const in a program, you still have to define it even if there's initializer in the declaration.

Checks wether Position is_inside, is_outside, or equals to other Position given the SentenceContext.