[FEA] Generalize collapse matchers to accept kv-dict and col list

[lmeyerov]

Is your feature request related to a problem? Please describe.

Typical cases of .collapse( are acting on a dict ({'type': 'account', 'risk': 'high'}) or a col list (cols=['type'] or cols=['type', 'risk']). Currently, a manual outer loop is required: for v in g._nodes[cols].drop_duplicates()... .

Describe the solution you'd like

g2 = g.collapse('root', match=['type', 'risk'])
g2 = g.collapse('root', match={'type': 'account', 'risk': 'high'})

It should also handle the case of the root id changing..

Describe alternatives you've considered

We can also support a general udf f(n1, n2), but until we get remote mode clearer, e.g., vectorization, wait

Additional context

It may make sense to wait until we've added collapse reductions and optimized those