…7240)

Both recommended intervals in the SDK docs exceeded our own free plan:

  EXAMPLES.md          time.sleep(300)  -> 288/day = 5.8x the 50/day allowance
  PERFORMANCE_GUIDE.md time.sleep(300)  -> same, and it is the "Solution" block

Defaults are now 1800 (30 min = 48 requests/day, fits free), with an inline
note to drop to 300 on Developer and above.

PERFORMANCE_GUIDE gains a plan/interval table and the measured update cadence
(BRENT ~2.5 min, WTI and natural gas ~5 min, refined products ~twice a day,
from the prices table over 7 days), so the interval is chosen from evidence
rather than habit. Above Developer, extra quota is better spent on more codes
than a shorter timer - the data has no more resolution to give.

Also flags api#7240 where a reader will trip over it: get_multiple() currently
issues one HTTP request PER CODE (prices.py:96-99 loops over self.get), so it
costs the same as a manual loop while the REST API accepts 20 codes in one
request that counts once. Verified against production today.

NOT changed after reading the context:
  PERFORMANCE_GUIDE.md:228 time.sleep(1) - explicitly labelled an anti-pattern

EXAMPLES.md was patched in binary to preserve its 640 CRLF line endings; a
formatter had otherwise normalised the whole file and turned an 8-line change
into a 1,286-line diff.

Refs api#7235, api#7240

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JKAExynd9zoKwt6rYA66EA