fix: make dbt-fal work with koldstart

[mederka]

    # HACK ALERT: This is a temporary solution for running Python models on Koldstart.
    # Environments in dbt-fal work with `isolate` abstractions and so far these
    # have been re-used to get us here. But `koldstart` API uses different
    # abstractions. So this where we tranform `isolate` objects into `koldstart`
    # objects and run the resulting isolated function.

    # One big change in `koldstart` is that it supports only one target environment
    # whereas isolate lets you stack environments on top of each other. We resolve this
    # in two ways, for `virtualenv` environments we just merge base requirements (dbt-*)
    # with user requirements, whereas for `conda` environments we create an `env_dict`,
    # where we add base requirements as `pip` requirements.

[linear[bot]]

FEA-889 Fix koldstart integration to dbt-fal

1. remove `isolate` completely 2. use `function.on(local)` when running locally