docs: adopt issue #4419 terminology in data sources guide

[andygrove]

Which issue does this PR close?

Part of #4419.

Rationale for this change

Issue #4419 proposes a terminology framework to remove the overloaded, bare use of "native" from Comet's docs. The data sources page is where the scan-type sense of "native" (a fully Rust-based scan versus a DSv2 / converted scan) is most prominent, and it carried the most bare "native" usages of any remaining user-guide page. Making the scan descriptions precise here is a natural follow-up to the Understanding Comet Plans and Scala/Java UDF cleanups.

This PR keeps the change scoped to that one page.

What changes are included in this PR?

In docs/source/user-guide/latest/datasources.md:

• Describe Rust-backed scans and readers as "Rust-based" or "runs in Rust via DataFusion" instead of "native" ("Parquet scans run in Rust", "Rust-based CSV scan support", "Rust-based reader", "Comet's Rust-based scans").
• Replace "allowing native execution to happen after that" with "allowing the Comet pipeline to take over after that" for the convert-to-Arrow passages.
• Use "offloads data loading to Rust" and "calling Java credential providers from Rust code" in place of "native code".
• Leave quoted plan/log output and operator names such as CometNativeScan verbatim.

How are these changes tested?

Documentation-only change. Verified with prettier --check.

[comphead]

Thanks @andygrove