Version bump and changelog for the #97 fix, so it can be published.

Why minor, not patch

The public signature is unchanged, so strict semver says patch. But this materially changes how many HTTP requests a call makes, and a minor keeps it out of tight pins like ~=1.12.8 while staying available to ~=1.12. Anyone on a narrow pin opts in rather than being surprised by a changed request pattern.

What ships

• get_multiple() batches — up to 20 codes per request instead of one per code. Through this method the free plan was 50 code-reads a day; through the raw API it is 1,000.
• Same fix in AsyncPricesResource.get_multiple(), which was worse — asyncio.gather fanned out one request per code, which could also trip the 60-per-60s rate limit.
• Polling examples default to an interval that fits the free plan, with the measured update cadence of the data.

Users' quota consumption drops. In their favour, no code change needed — but it is a behaviour change and the changelog says so plainly.

One test fixed along the way

test_package_version_helper_reads_the_project_version asserted the literal "1.12.8", so it failed on this release for a reason unrelated to what it guards. Its job is "does the helper report the project version", not "is the version 1.12.8". It now reads the expected value from pyproject.toml — so it keeps testing the helper and stops needing a hand-edit every release.

Verification

645 passed, 42 skipped, 0 failed
coverage 63.44% (gate 50%)
import oilpriceapi; oilpriceapi.__version__ -> 1.13.0

Live smoke on the merged fix, against production:

 10 codes -> 1 request,  10 prices   (was 10)
 30 codes -> 2 requests, 29 prices   (was 30)
 bad code -> per-code fallback, 1 ok + 1 reported failure

Publishing is triggered by a published GitHub Release, so merging this does not ship to PyPI on its own.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JKAExynd9zoKwt6rYA66EA