fix(pricing): page through the whole catalog in pricing sync

[andreacappelletti97]

Bug

The documented air-gapped workflow didn't work. c3x pricing sync reported "Offline ready", but c3x estimate --offline against the synced cache priced an m5.xlarge at $0 (the stub).

Cause: the product enumerator requests pages of 5,000. It treated any page shorter than that as the last one, and stepped the offset by the requested size. c3x-pricing-api#60 capped requests at 1,000 products (MAX_PRODUCTS_PER_REQUEST), so every page comes back short and the sync stopped after the first 1,000 products of every service and region. The live API confirms this: asking for limit:5000 returns 1000.

Fix

• Advance the offset by the rows actually returned, and stop only on an empty page. A server cap of any size can no longer truncate a sync.
• Regression test: a stub server capped at 3 per page with 7 products. The old code saw 3; the new code sees all 7.

Evidence

Live pricing.c3x.dev, pricing sync --providers aws --regions us-east-1, then the same config priced offline and live:

	before	after
prices synced	4,455	22,364
cache size	1.6 MB	9.1 MB
offline m5.xlarge	$0 (stub)	$140.16
offline total vs live	$11.50 vs $334.16	$334.16 = $334.16

Released CLIs still have the old paging. c3x-pricing-api is raising its default cap to 5,000 in a separate PR, so their one-page requests are complete again. That also cuts the sync time (7m56s at 1,000 per page).