Open .conda ZipFile once per extract (B13)

[jezdez]

Description

cph.streaming._extract iterates over the two components (pkg and info) of a .conda archive, calling cps.stream_conda_component twice. Each call internally instantiates a fresh zipfile.ZipFile(fileobj), parsing the central-directory end-of-file record from scratch.

This PR opens the ZipFile once in _extract and threads it through both stream_conda_component calls via the new zf= kwarg added in the companion cps PR (conda/conda-package-streaming#173). No semantic change; saves one ZIP-parse per archive.

Depends on conda/conda-package-streaming#173 landing first — without the cps-side zf= kwarg this is a no-op.

Part of the conda-tempo Track B performance work. Tracking issue: conda/conda#15969.

Phase-2 micro-benchmark (bench_s13_zipfile_single.py, 10 .conda archives, both components):

	Before	After	Speedup
10 archives, 2 components each	999 µs	502 µs	2×

Phase-4 end-to-end (as part of the full cps + cph stack):

Workload	mac baseline → stacked	Linux baseline → stacked
W4 (cold-cache data-sci install)	43.88 s → 36.14 s	26.28 s → 23.38 s

Full research report: https://github.com/jezdez/conda-tempo/blob/main/track-b-transaction.md

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests? (existing extract tests cover the .conda → pkg + info round-trip; happy to add a regression that asserts only one ZipFile is constructed)
• Add / update outdated documentation? (no user-facing docs change)

[dholth]

@jezdez this can be continued now that conda-package-streaming has been updated