Can we use the arrow PyCapsule/C api when ingesting from pyarrow.Table?

[timkpaine]

we have a relatively long-winded ingestion from a pyarrow.Table:

perspective/rust/perspective-python/src/client/python.rs

Line 166 in 768ca40

fn to_arrow_bytes<'py>(

Even though it would start as a different format, can we save time by going through the pycapsule/c interface: https://arrow.apache.org/docs/format/CDataInterface/PyCapsuleInterface.html#consume-a-pycapsule

xref: #1157

[texodus]

Converting to a discussion - definitely interested in higher performance Arrow ingestion!