Define and expose sequence-related concepts

[tturocy]

This is the first phase of work towards the objective of #494, migrating common sequence form-related concepts into
the game representation, and making these available in pygambit.

This does the following:

• Sequences are now defined on the game and generated just once (similar to how strategies are done). Previously, enumpoly would generate sequences for each support separately.
• enumpoly uses these game-level sequences (by filtering the subset of them which are relevant given the support under consideration)
• In C++, the tree structure of sequences (for each player) is encoded via the parent member.
• Creates sequences as a concept in pygambit (again parallel to the treatment of reduced strategies). A small number of tests are added to exercise this.

To review this, we should consider:

• Are the tests adequate enough and well-organised? (for this initial draft probably not).
• Are the docstrings sufficient and are all the relevant functions added to the documentation?
• Do we want to extend any of the user documentation to make reference to these?

What is outside of the scope of this PR will be any population of a payoff tensor or constraint matrix; these remain for the moment the province of the separate algorithms that use the sequence form representation (this will be addressed in subsequent stages).