Problem: lack of indexing primitives

[yrashk]

Even though indexing primitives are promised in our documentation, we haven't published them yet, leaving this to users.

Proposed solution: before addressing particular index types (equality, ordered, etc.), design a common approach to defining indices. Aspects that I can think of:

• How explicit every occurrence of indexing is? Should we require manual indexing on every recording or should the indexing subsystem "know" how to index a particular key?
• How do we deal with indexing values that were already added?
• How do we deal with pausing and resuming indexing?

It's likely that the system can be designed in a layered fashion with bare indexing primitives at the bottom and more high level functionality resting atop of it.

This is very likely to end up being an RFC.

[yrashk]

I was recently discussing this topic with @stuarth and we might need to remove this from the 0.2 milestone (as in "adding indexing primitives to PumpkinDB in a form of code") and write a chapter on indexing methods in the documentation instead, until we can figure out the proper API and (possibly) write an RFC. The documentation itself should probably be targeted for 0.2 anyway (in this case, the issue can stay assigned to that milestone)