fix(android): remove dead VaultWebSocketClient stub (#178)

[yahia008]

Title:
fix(android): remove dead VaultWebSocketClient stub duplicating VaultEventSocket (#178)

Body:

Summary

Closes #178.

VaultWebSocketClient.kt was a complete, correctly-structured WebSocket client whose
own doc comment called it a stub for #110 and said to "wire up the Flow into
VaultViewModel once [a staging endpoint] does." That wiring already happened — under
a different class. VaultEventSocket.kt is the real, @Singleton-injected client
VaultViewModel actually depends on, complete with reconnect backoff and its own test
suite. VaultWebSocketClient had zero references anywhere in src/main, src/test,
or src/androidTest — never instantiated, never bound in AppModule.kt. Leaving it in
the tree risked a future contributor building against the wrong, unmaintained client.

Changes

• Deleted android/app/src/main/java/com/ethosprotocol/api/VaultWebSocketClient.kt
  (186 lines), including its VaultWebSocketEvent sealed class and the private
  message-envelope DTOs (WSVaultUpdatedMessage, WSVaultExpiredMessage,
  other file references any symbol from it.
• Confirmed no event coverage was lost: VaultEvent (in models/Models.kt) already
  covers vault_updated / vault_expired / vault_released generically via its type
  discriminator + optional vault field — nothing needed to be ported from the stub.

Explicitly out of scope

• VaultEventSocket.kt's runtime behavior is untouched — this PR only removes dead code.
• iOS has an identical dead-stub twin (ios/.../Services/VaultWebSocketClient.swift)
  alongside its own real, wired VaultEventSocket.swift. Same pattern, but out of scope
  for this Android-scoped issue — worth a follow-up issue if desired.

Test plan

• Repo-wide grep -rl "VaultWebSocketClient\|VaultWebSocketEvent" returns no hits
  outside the deleted file (verified locally before commit).

• Added CI step in android-ci.yml asserting the same, so reintroduction fails CI.

• CI (./gradlew assembleDebug, testDebugUnitTest, verifyPaparazziDebug,
  assembleRelease, lint) — pending on this PR since no local JDK/Android SDK was
  available to run Gradle directly in the dev environment used to prepare this change.

  • Added a CI guard in .github/workflows/android-ci.yml (build-and-test job): a
    grep step that fails the build if VaultWebSocketClient or VaultWebSocketEvent is
    ever reintroduced, so this can't quietly come back.
  • Fixed PARITY.md: the "Live vault updates via WebSocket" row was stale (❌/❌,
    "Neither client implements WebSocket yet") despite VaultEventSocket being real and
    wired on both iOS and Android — updated to ✅/✅, and removed the matching stale entry
    from the "Known gaps" table.
  • Checked VaultEventSocket.kt, VaultViewModel (ui/ViewModels.kt), and
    shared/api-contract.md for #110/stub cross-references pointing at the deleted
    file — found none, so no further edits were needed there.

  Explicitly out of scope

  • VaultEventSocket.kt's runtime behavior is untouched — this PR only removes dead code.
  • iOS has an identical dead-stub twin (ios/.../Services/VaultWebSocketClient.swift)
    alongside its own real, wired VaultEventSocket.swift. Same pattern, but out of scope
    for this Android-scoped issue — worth a follow-up issue if desired.

  Test plan

  • Repo-wide grep -rl "VaultWebSocketClient\|VaultWebSocketEvent" returns no hits
    outside the deleted file (verified locally before commit).
  • Added CI step in android-ci.yml asserting the same, so reintroduction fails CI.
  • CI (./gradlew assembleDebug, testDebugUnitTest, `verifyP