multiple levels of blueprint verification

The design philosophy behind all of the curent conduit::blueprint::verify functions has been to approach the problem of schema validation from a data-agnostic perspective. While this makes validation straightforward and efficient, it excludes a number of more thorough checks that could be performed to more confidently report the validity of a given Blueprint hierarchy (e.g. entries in the topology and adjsets section could have their index values bounds-checked against the length of the associated coordset), which will become more relevant as the data becomes more complex (see #511, #513). These checks could be included conditionally based on user preferences, which would require adding a parameter to the conduit::blueprint::verify function to indicate a "level" of thoroughness.

Start by reading the existing conduit::blueprint::verify functions and the related discussions in #511 and #513. Define the verification levels and the scope of checks such as topology and adjsets index bounds before deciding how the API should expose them; done means the level semantics and affected checks are agreed.