Advertise byte ranges on downloads, and record the live checks

Stalwart honors a single byte range on its download endpoint but sends
no Accept-Ranges, and Chrome's PDF viewer only reads a file in pieces
when the first response says it can. The proxy now says so itself.

Checked live on 0.16.22: ContactCard/changes reports creates, updates
and destroys exactly, which the contacts store's sync relies on, and a
range the server cannot serve gets the whole file with 200, never 416.
The mock now answers ranges the same way and sends no Accept-Ranges.

KNOWN-ISSUES.md

@@ -48,6 +48,8 @@ works the same way — and dropped where 0.15 was the whole subject. Support for
 0.15 was removed on 2026-08-26; the last release that runs on it is tagged
 [`stalwart-0.15-support`](https://github.com/Coffey-Labs/ihasmail/releases/tag/stalwart-0.15-support).
 
+- **`ContactCard/changes` works, and a download honors one byte range but does not say so.** Both **confirmed live (0.16.22, 2026-09-16)**, with objects on a throwaway account that were removed afterwards. `ContactCard/changes` reports a create, an update and a destroy exactly, nets a card created and destroyed since the given state out to nothing, and answers a state it does not recognize with `invalidArguments` rather than `cannotCalculateChanges`; the contacts store syncs from it and falls back to a full reload on any error. The download endpoint answers a single range (`bytes=0-9`, `bytes=-5`, `bytes=995-`) with `206` and a correct `Content-Range`, and anything else (several ranges, or a range past the end) with the whole file and `200`, never `416`. It sends no `Accept-Ranges`, so ihasmail's proxy advertises it: Chrome's PDF viewer reads a file in pieces only when told it can. The mock answers the same way.
+
 - **Push subscriptions are not replaced by a repeated `deviceClientId`, and an account holds fifteen.** ihasmail registered a new subscription on every renewal believing the old one would be replaced, as the mock did. **Confirmed live (0.16.22, 2026-09-16)**: a second create with the same `deviceClientId` leaves both in place, the sixteenth create is refused with `overQuota`, "There are too many subscriptions, please delete some before adding a new one.", and `update` of `expires` is accepted. `PushSubscription/get` does not return `url` (nor `keys`), so a subscription can only be matched by its `deviceClientId`. A `types` of `[]` or `null` is stored as *every* type, not none. Read from the 0.16.22 source: `EmailDelivery` changes only on delivery, a delivery reaches a subscription with an `emailPush` filter as an EmailPush alone, and the payload carries `id` and `threadId` only when they are named in `properties`. Browsers now subscribe to `EmailDelivery` only, extend rather than re-create, clear their own duplicates and make room on `overQuota`; the server removes what its previous process registered. The mock follows all of it ([#375](https://github.com/Coffey-Labs/ihasmail/issues/375)).
 
 - **A contact photo has to be a `data:` URI; Stalwart refuses one given as a `blobId`.** RFC 9610 lets JMAP put a `blobId` in a JSContact `Media` object, and ihasmail uploaded the photo and saved it that way, which the mock accepted. Stalwart does not: **confirmed live (0.16.22, 2026-09-16)**, a `ContactCard/set` create with `media.*.blobId` fails with `invalidProperties` on `media`, "blobIds in media is not supported." The RFC 9553 `uri` form with a `data:image/jpeg;base64,…` value is accepted on create and on update, and `ContactCard/get` returns it unchanged; a 134 KB one was accepted. Photos are now saved inline, and the mock refuses a `blobId` the same way ([#376](https://github.com/Coffey-Labs/ihasmail/issues/376)).