Data type consistency between `to_iter`/`to_list` and `read_records`

[ai-ignatyev]

Description

Let's assume we have a simple csv file with a boolean field:

x
false
true

If we try to read this cvs file, and then recreate the chain with to_iter/read_records, we encounter with an error:

import datachain as dc

chain = dc.read_csv('sample.csv')

records = [
    {k: item[i] for i, k in enumerate(chain.schema.keys())}
        for item in chain.to_iter()
]

dc.read_records(records, schema=chain.schema)

Error while validating/converting type for column x with value 0, original error Value 0 with type <class 'int'> incompatible for column type Boolean

This happens due to to_iter returns int despite bool in the schema, but read_records requires strict type matching. So, I think it would be more convenient if to_iter returned bool or read_records could parse int.