Add option for lazily evaluating data

[mallport]

This change makes it so that the from_polars() method can accept a Polars LazyFrame. Only the columns that are pseudonymized should be materialized, the rest is chunked and streamed directly from bucket-to-bucket

This change is 

[mallport]

After we upgraded Polars, we no longer have to use file handles for writing to GCS, we can just supply the "gs://"-filepath and Polars automatically infers the authentication from the environment

[skykanin]

@skykanin made 4 comments.
Reviewable status: 0 of 19 files reviewed, 5 unresolved discussions (waiting on mallport).

---

src/dapla_pseudo/v1/pseudo.py line 122 at r2 (raw file):

            if hierarchical and type(Pseudonymize.dataset) is pl.LazyFrame:
                raise ValueError(
                    "Hierarchical datasets are not supported for Polars LazyFrames."

We should probably note this in the dapla manual

---

src/dapla_pseudo/v1/result.py line 157 at r2 (raw file):

                return pandas_df
            case pl.LazyFrame() as ldf:
                pandas_df = ldf.collect().to_pandas()

.collect realises the lazyframe as a dataframe. This might be a footgun for some users and it's probably never what you want if you're using a lazyframe to begin with. We should at least document this behaviour in the pydoc comment for to_pandas or decide not to support this case and throw an exception

---

tests/v1/integration/test_pseudonymize.py line 152 at r2 (raw file):

@pytest.mark.usefixtures("setup")
@integration_test()

Would it be possible to add a regression test to ensure we're using an expected amount of memory when loading large datasets as a lazyframe and pseudonymizing a few columns?

[skykanin]

@skykanin reviewed 21 files and all commit messages, made 2 comments, and resolved 2 discussions.
Reviewable status: 21 of 22 files reviewed, 5 unresolved discussions (waiting on mallport).

---

tests/v1/integration/conftest.py line 20 at r3 (raw file):

        # Need to disable local file logging to avoid getting gcloud perm error
        # subprocess.run(["gcloud", "config", "set", "core/disable_file_logging", "True"])

This should be uncommented again

---

tests/v1/integration/test_lazy_regression.py line 53 at r4 (raw file):

@pytest.mark.usefixtures("setup")
@integration_test()
def test_lazy_projection_memory_regression() -> None:

We should test the difference in memory usage between loading and pseudonymizing the same data in a lazyframe vs a regular dataframe

[skykanin]

@skykanin reviewed 2 files and all commit messages, and made 1 comment.
Reviewable status: all files reviewed, 5 unresolved discussions (waiting on mallport).