docs(benchmarks): slim CAMPAIGN.md's finished checklist rows and history #360

benchmarks/experiments/CAMPAIGN.md has grown to about 122 KB, and about 72 KB of it is the section 7 re-pin checklist. Its own rule says a row that is built and verified "is condensed to a line that keeps its number", yet about 66 rows marked built, done or decided average about 0.9 KB each, with the implementation detail of the change duplicated from the code, tests and DECISIONS/BUGS entries they already point to.

Proposal, docs only, one PR: condense every built, done or decided row of the checklist to its number, one sentence on what changed and the pointer it already has (commit, file, test, DECISIONS/BUGS citation), keeping open rows and rows that carry an instruction for a future run whole. No row is renumbered, reordered or deleted, because code, tests and other docs cite rows by number. Also cut the "Why Phase B is not parallel" and "Why the build/query phase split is not adopted" sections to a short rule each, the history paragraph under "ONE LINE OF DEVELOPMENT" to its rule, the dated owner quotes and stories in section 8 to the numbered steps, and the dated "last swept" and "started" remarks; reword the mentions of upstream_ledger.py and upstream_watch.py, which are private notes tools and not in this repository.

Text that a check parses (the October chain paragraph read by structure_check.py, rows 70 and 73 read by tests) stays as it is. Verification: structure_check.py, the benchmark tests that read CAMPAIGN.md, and pre-commit.