fix(notifications): route push transports by explicit provider

[eleboucher]

Description

Fixes several Android push-notification bugs, and bumps tauri-plugin-notifications to pick up the matching native fixes (SableClient/tauri-plugin-notifications#11).

Transport routing

• Pass provider: 'fcm' / provider: 'unifiedpush' explicitly on register, so the plugin routes deterministically instead of inferring from saved state (switching to native no longer keeps using the old UnifiedPush distributor).
• switchBackgroundPushTransport is break-before-make (deactivate old → register new), which the plugin's single-active-provider guard requires. If the new registration fails, the previous transport is now restored instead of leaving the device with no pusher while the UI still says push is on.
• Skip no-op mode changes when the resolved provider is unchanged.
• enablePushNotifications no-ops under Tauri instead of throwing (fixes the SABLE-162 unhandled rejection).

Background notification previews

• Encrypted message bodies could appear in notifications with "show encrypted content" disabled: isEncryptedRoom is derived from the local room, which is absent on a cold start from push, so the gate passed and the decrypted body was shown. Encryption is now taken from the fetched event itself.
• A notification dismissed while its preview fetch was in flight could reappear ("ghost" notification), the async update now checks the cache is still the live one.
• An earlier message no longer stays stuck on the "New message" placeholder when a second push arrives before its preview resolves.
• The second (enriching) post is awaited so failures are logged instead of becoming unhandled rejections, and the per-room seen-event set is capped.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings

AI disclosure:

• Partially AI assisted (clarify which code was AI assisted and briefly explain what it does).
• Fully AI generated (explain what all the generated code does in moderate detail).