feat(stream-android): add Sendbird → Stream Chat migration skill

[aleksandar-apostolov]

Adds a Sendbird → Stream Chat migration skill for Android (Kotlin / Jetpack Compose / XML Views) — a repeatable, empirically-hardened procedure for migrating any Sendbird integration to Stream. Sibling to the iOS migration skill, but built from the Android SDK's actual surface and validated by end-to-end dogfooding, not translated from iOS.

What's here

• skills/stream-android/sendbird-migration.md — the skill: detect the integration shape (Sendbird uikit-compose / XML uikit / custom UI on the low-level sendbird-chat SDK / spaghetti; MVVM/MVI; single- vs multi-activity; Compose-Nav vs fragments+Jetpack-Nav), pick one of three paths (Stream Compose bundled screens / keep-the-UI-swap-the-SDK surgical / cross-framework XML→Compose via ComposeView), swap packages + init/auth, map channels/messages/state, re-apply theming + custom messages, and verify fidelity per screen.
• SKILL.md — a migration flag routes "migrate off Sendbird" requests to the skill (orthogonal to the A/B/C/D tracks).

Load-bearing pieces

• Gap-handling protocol for screens Sendbird bundled that Stream doesn't: detect the dropped screen, then either wire the bundled Stream equivalent (GroupChannelInfoScreen/DirectChannelInfoScreen) or — for screens Stream doesn't ship (channel creation) — tell the user it isn't out-of-the-box, link the public cookbook, and offer to build it. Never silently drop or invent a screen.
• v7 packaging gotcha — the state layer is folded into stream-chat-android-client (io.getstream.chat.android.client.api.state), state/offline aren't published separately, and the client auto-installs state (no withPlugins). Verify against the published artifact, not the dev tree.
• State-layer mapping (collections → queryChannelsAsState/watchChannelAsState), custom-message carriers (Message.extraData vs a custom AttachmentFactory), and the always-needed solid-bubble + white-text and DM-name fidelity fixes.

Built + validated empirically

• Grounded in real migrations of Sendbird source apps across architectures, then dogfood-tested: a fresh agent given only the skill produced a correct migration plan including the gap-handling, and a separate fresh agent migrated a spaghetti / low-level-SDK app end-to-end to a first-try successful build that connects and renders.