Added support for signatures evaluation/tracking and added to tokenizers collection.
Changed how the sidecard provides data (now are proper JSONs) and unified the code with LMEH functions used for tokenizer loading.
Updated the protocol.py classes adding results and tokenizers database entries.
Added two collections to hold buffers: buffer_numerical and buffers_signatures. These were removed from nodes collection.
Added deletion of entries in tasks, instances, results, responses and prompts collections after a successful processing from the Manager.
Added configuration override to allow a number of tasks to trigger even if the number of samples in the buffer is filled (required by the signatures:tokenizer tasks).
Fixed some issues with circular buffer calculation.
Updated the numerical task response format in the Evaluator.
Two issues solved in this PR ( https://github.com/pokt-scan/pocket-ml-testbench/issues/41 and https://github.com/pokt-scan/pocket-ml-testbench/issues/64 ) . Also this PR includes many minor fixes to complete a successful generation of scores from an empty DB:
tokenizerscollection.sidecardprovides data (now are proper JSONs) and unified the code with LMEH functions used for tokenizer loading.protocol.pyclasses addingresultsandtokenizersdatabase entries.buffer_numericalandbuffers_signatures. These were removed fromnodescollection.tasks,instances,results,responsesandpromptscollections after a successful processing from the Manager.signatures:tokenizertasks).