Improve Synthetic test coverage

[nyurik]

Our synthetic test cases generated by SyntheticMltGenerator.java are not covering some cases. Feel free to implement any one of these, run just generate-synthetic-mlts and create a PR.

• u8 property columns
• i8 property columns
• u32 property columns feat(java): support UInt32 and UInt64 encoding/decoding for synthetics #852
• u64 property columns feat(java): support UInt32 and UInt64 encoding/decoding for synthetics #852
• Double.MAX_VALUE cannot be encoded
• morton-encoded data
• pre-tesselated data - chore(java): add tesselation synthetics #837
• Optional IDs (both i32 and i64)
  • vibed PR needs feedback: fix(java): update Feature ID handling to support nullable IDs #829
• Optional for every possible property type
• Go through every possible encoding and ensure each test case exists
  • integers
  • floats test: To infinity and beyond! (adding prop_{f32,f64}_{min_inf,max_inf,nan} synthetics) #858
  • strings test: To infinity and beyond! (adding prop_{f32,f64}_{min_inf,max_inf,nan} synthetics) #858
• Shared dictionaries test: Add shared dictionary synthetics #854
• Coordinates inside and outside the default tile extent of 0..4095, e.g. -42 or 4112 - chore: add extent tests #853
• Non-default extent size - e.g. 512 or 2^20. We should be able to support coordinate representation of i32 -- everything up to 2^30 (buffer is important, so extent=2^31 would allow negative buffer, but not the coordinates with beyond 2^31) - chore: add extent tests #853