Incremental encoding/decoding slows down string value parsing for fast field.

[ChangRui-Ryan]

Is your feature request related to a problem? Please describe.
As I mentioned in my pr: #2795, The str-type fast fields in Tantivy adopt incremental encoding, which means retrieving each original string requires decoding from scratch—resulting in extremely low efficiency and sluggish speed. This also impacts the processing of str-type values based on TopDocs, especially when the limit is relatively large.

Describe the solution you'd like
Is it possible to choose not to use incremental encoding and only adopt basic dictionary encoding via parameter configuration, so that the original string value can be directly retrieved from the dictionary encoding of the str-type fast field? This approach is more user-friendly in many scenarios.

[fulmicoton]

@ChangRui-Ryan Yes, tantivy assumes that you can do all of the heavy lifting in term ordinal space. Can you describe the actual scenario you are trying to solve?

[ChangRui-Ryan]

@fulmicoton Sorry for the late reply. We use Tantivy as our storage engine and frequently read plenty of field values from it (based on search conditions); Str is also a field value type, but its read performance is much worse than INT and other similar types.