Implement an "autotuner" for encoder settings

[Ostrzyciel]

Jelly has quite a lot of knobs to turn and adjust, with their amount only expected to increase:

• Everything in RdfStreamOptions
• Which prefixes, names, and datatypes are most common – we can place them in low-ID areas of lookup tables and protect them from eviction
• (future Investigation: text compression in literals jelly-protobuf#16) Which binary compression algorithm to use, and at which level (note that this matters a lot for things like brotli)
• (future Proposal: support embedded protos for literal serialization jelly-protobuf#14) Which datatypes would it make sense to encode in the specialized binary format
• (future) Whether or not would it make sense to use diff-based, frame-to-frame compression, and which parts of the frame should be in the unchanging baseline.

Knowing which settings to use boils down to investigating the data at hand. Could we implement some simple heuristics that do this automatically?

The results of the analysis should be saved in a case class instance that could be directly passed to the encoder, initializing it. I think it would be best if we settled on this mechanism for encoder initialization, because it will make it easier to add new features and options, without changing factory method signatures.

Maybe we could also allow saving these settings as a proto? I'll make a follow-up task for this. Jelly-RDF/jelly-protobuf#17

[Ostrzyciel]

The caller should also be able to specify additional information about the use case/intended channel. For example, are we interested in maximizing the serialization throughput, deserialization throughput, or representation size?