data type discrepancies in choose_array_encoder function

[broccoliSpicy]

currently, there are many data type discrepancies in choose_array_encoder which makes encoding selection logic harder and bug-prone to implement, for example:

in this code, the array's type is still one of BINARY_TYPES but the data_type parameter of choose_array_encoder has changed to UInt8

lance/rust/lance-encoding/src/encoder.rs

Lines 303 to 314 in f48e970

	else if BINARY_DATATYPES.contains(arrays[0].data_type()) {
	if let Some(byte_width) = check_fixed_size_encoding(arrays, version) {
	// use FixedSizeBinaryEncoder
	let bytes_encoder = Self::choose_array_encoder(
	arrays,
	&DataType::UInt8,
	data_size,
	false,
	version,
	None,
	)?;

same thing happened here:

lance/rust/lance-encoding/src/encoder.rs

Lines 223 to 242 in f48e970

	fn default_binary_encoder(
	arrays: &[ArrayRef],
	data_type: &DataType,
	field_meta: Option<&HashMap<String, String>>,
	data_size: u64,
	version: LanceFileVersion,
	) -> Result<Box<dyn ArrayEncoder>> {
	let bin_indices_encoder =
	Self::choose_array_encoder(arrays, &DataType::UInt64, data_size, false, version, None)?;
	let compression = field_meta.and_then(Self::get_field_compression);
	let bin_encoder = Box::new(BinaryEncoder::new(bin_indices_encoder, compression));
	if compression.is_none() && Self::can_use_fsst(data_type, data_size, version) {
	Ok(Box::new(FsstArrayEncoder::new(bin_encoder)))
	} else {
	Ok(bin_encoder)
	}
	}

we need to have some coherence for encoding selection predicate