Standardize JanusGraph access pattern throughout doc code examples

I propose that we go through the documentation and find instances of JanusGraph being used in embedded mode and switch these over to a remote access example via the TinkerPop driver. Then add a section near the beginning of the docs that outlines the reasons for and against running JanusGraph in embedded mode.

Search the documentation for examples that use JanusGraph in embedded mode, then identify the corresponding remote-access examples using the TinkerPop driver. Review the documentation's early sections for a suitable place to explain the trade-offs of embedded mode; done means the examples are standardized and the rationale section covers both advantages and disadvantages.

Written by the indexing model from the issue text.