Allow tuples/iterables as convenience for setting contraints on model parameters

[embray]

Currently it is possible to specify constraints on parameters when instantiating a model by providing keyword arguments named after the constraints, and dict that provides settings to that constraint for some (or all) parameters. For example:

m = SomeModel(1, 2, 3, fixed={'param_a': True, 'param_b': False})

The dict-based assignment is definitely nice and explicit, and also good if you only need to specify some constraint on a few parameters. But as @mhvk suggested it would also be convenient to accept a sequence that provides that constraint for all parameters. For example if the parameters on a model are 'param_a', 'param_b', and 'param_c' one could write:

m = SomeModel(1, 2, 3, fixed=(True, False, True))

instead of

m = SomeModel(1, 2, 3, fixed={'param_a': True, 'param_b': False, 'param_c': True})

This would work since by design the parameters on each model have a standard "order" associated with them. I see no harm in allowing both styles.

[astrofrog]

@embray, since this is marked as low priority, shall we remove the 1.0 milestone?

[Lyalpha]

Would the same behaviour be desirable for the bound and tied kwargs, or should it be restricted only to fixed?

[embray]

There's no reason it should be just for any one constraint, no.

[embray]

In fact at some point I plan to generalize the constraint handling code, since one thing that came out of #pyastro15 was the (eventual) need for more types of parameter constraints.