Outstanding improvements for the encoder and decoder

[mactrem]

Several components require further development to transition MLT from Research-Grade to Production-Grade.

Java Encoder

• Efficient encoding selection: Implement fast, sample-based encoding selection algorithms as proposed by BTRBlocks, replacing the current brute-force approach
• Adaptive dictionary selection: Develop a proper selection strategy for Shared Dictionary and Shared FSST Dictionary Encoding, instead of relying on hardcoded column specifications

JS Decoder

• Enhanced FSST decoder: Integrate an optimized version of the FSST decoder to improve performance
• FastPFOR integration: Add support for a FastPFOR-based decoder to enable high-performance decoding of compressed integer sequences