Avoid abuse dictionary encoding in parquet writer

[ShiKaiWi]

Describe This Problem

Currently, the default configuration is used for parquet writer to generate the sst, and dictionary encoding is used for bytes column (including string column) by default. Only the dictionary page exceeds 1MB, the parquet writer will fall back to plain encoding without the dictionary.

However, it maybe not very efficient for the column of high cardinality to be encoded by dictionary.

Proposal

Maybe we should choose to control whether to choose dictionary encoding according to the column's cardinality. However, the threshold of cardinality to enable dictionary encoding should be benchmarked and tested.

Additional Context

No response