JPEG2000 input transformation

[treigerm]

Raising this issue so that we remember the discussion from #15 .

Currently, given an input x we are dividing it by the absolute error bound eb. We should ideally ensure that this division does not bring us outside of the range of the data that can be represented with a signed integer (25-bit signed integer if I remember correctly?).

compressor/src/climatebenchpress/compressor/compressors/jpeg2000.py

Lines 33 to 44 in 2b3d895

	return CodecStack(
	numcodecs_wasm_fixed_offset_scale.FixedOffsetScale(
	offset=0,
	scale=precision,
	),
	numcodecs_wasm_round.Round(precision=1),
	numcodecs.astype.AsType(
	encode_dtype="int32",
	decode_dtype="float32",
	),
	numcodecs_wasm_jpeg2000.Jpeg2000(mode="psnr", psnr=psnr),
	)