Add support for closing with dependent lambdas on types

[umazalakain]

We might want to allow dependent lambdas binding types in the future, changing to a data type identifiers seems like a hack we should get rid of.
Initially things blew up because some traversal paths for dependent lambdas on type identifiers were missing, once I added those the trouble seemed to be with the fission tests in the fissionFussion suite and the simple fusion test in the traversals suite: they use mapLastFission, which only works on data types, weakening this condition breaks inference on the rewritten term.
Originally posted by @Bastacyclop in #176 (comment)