Restores the settled block's bullet indent

[johnnyt]

What

Two fixes to the "Flagged, not settled: the provider is consulted on every call (settled 2026-09-13)" comment section in Encryptor.Vault.Encrypt, plus one dated foot Note on ADR-0001.

• The section's two record bullets get their hanging indent back: two-space bullets with four-space continuation, the style this file used before the block was rewritten and the style Encryptor.Vault.Suspension's comment bullets use. Only leading whitespace changes on those lines. The block still starts with its heading at :57 and ends at :84, the range and heading string ADR-0002 Amendment A cites.
• The section quotes the vault's ordering rule ("resolves the selector through the provider, validates and maps the descriptor to a keyring, composes the context, derives the partition id, builds the CMM stack and the client"), but no record contained the sentence; ADR-0002 Amendment A's A1 carries only its first clause. A new Note at the foot of ADR-0001 records the sentence verbatim, names the bead it came from, and cites by anchor where the encrypt path follows each step. The comment's attribution now points at that Note, rewrapped inside the same seven lines, so the block's range holds.

Comment and record only: no behaviour changes and no changelog fragment (changelog.d/README.md excludes documentation and ADRs).

Provenance

• The acceptance allowed either an ADR-0001 cite by anchor and SHA in the comment, or the sentence recorded in ADR-0001 by a dated foot Note. I chose the Note because ADR-0001 does not contain the quoted words, so a cite would leave the quotation without a source. The comment's pointer to the Note is the change the bead requires to link the two.
• The Note says the partition id is derived only when a cache is configured, because maybe_caching/3 calls Partition.id/2 and the cache: false clause skips it. It reports what the code does and decides nothing.

Review (in-turn, gate tier)

I read the diff against the bead's two acceptance items and checked every anchor in the Note against lib/ at 827c6d3, the base of this branch; lib/ is byte-identical between 32234d6 and that commit. Resolve.encryption_key/3 is called inside the provider span at encrypt.ex:141, Keyring.build/3 at :143, and Resolve.context/5 at :144. Keyring.build/3's catch-all refusal is at keyring.ex:120, and Partition.id/2 is called inside maybe_caching/3 at encrypt.ex:192. client/3 is defined at :173 and stack/3 at :160. The edit adds and removes no lines in encrypt.ex, so every anchor stays put, including ADR-0001's own cites of encrypt.ex:59-84, :162-164 and :184-194. git diff origin/main -- docs/adr/ shows zero removed lines, and the Note sits after the file's last heading. mix quality is green on the committed tree (673 tests, Credo, Docs, doc links, Dialyzer). The gate lock and a machine slot were held for that run, and the staged tree is byte-identical to the tree that ran, so the commit was made without a second run.