Separation of state between ShexValidator and ValidationContext

[ericprud]

The life-cycle of a ShexValidator and the ability of users to supply specialized ValidationContexts will dictate which attributes appear in a validator vs a context.

currently:

ShexValidator:

    Collection<SemanticActionPlugin> semanticActionPlugins

ValidationContext:

    private final ShexSchema schema;
    private final Graph data;
    private ValidationContext parentCtx = null;
    private final Map<String, SemanticActionPlugin> semActPluginIndex;
    private final SorbeFactory sorbeFactory;
    private final Deque<ValidationStackElement> validationStack;

ShExMap needs access to the siblings and the arities of all expressions on the stack when any variable was bound.

Iovka and I discussed moving all state from the validator and making it static. Given that ShExMap needs access to the hierarchy of the AST (including cardinalities compiled out into TripleExprCardinalities), and that detailed error reporting could also benefit from such positioning, I'm currently tempted to move schema, data, and sorbeFactory from the context to the validator.