✨ [DRAKEN] TO DO

[joocer]

• Write the InList and Range searches
• Design the API, we should provide results back in the form: { "{filename}": [rows] }
• Write AccumulatorMemtable, which instead of tracking the last entry for a key, we accumulate rows
• Add in some thresholds
  • more than 100k keys don't use the BloomFilter (this is 10% FP rate), the searches should not use the BF if it's zero bytes
  • more than a threshold (1000 rows?), don't collect any more rows - hint to the engine that it's not specific enough to be useful
• Write the Python parts of the API (the FastAPI and storage access bits)
• Write the Client part, that will be used by Opteryx and Mabel (Opteryx should also be able to build an index)
• Exemplar use-case will be asset searching (about 50k unique values in a 3m row dataset)