if the server is unable to query the database throwing a temporary-auth-failure
might be more appropriate

* fix-contact-sync:
  Contact sync needs to happen when we discover a new gateway
  Skip contact sync work if there are no gateways configured

* 'master' of github.com:iNPUTmice/Conversations:
  bump agp
  version bump to 2.10.6
  pulled translations from transifex
  do not merge failed decryptions
  version bump firebasse-messaging lib

Since that can change the outcome of the sync operation.

Or if syncing before gateways have been discovered, etc.  Don't do the work if
we know nothing will come of it.

* darker:
  Darker Obsidian

* 3GPP-video:
  Move video MIME above audio MIME

So that when a file extension is ambiguous, we guess video.

fixes #4314

* 'master' of github.com:iNPUTmice/Conversations:
  pulled translations from transifex
  Revert "flush stanzas in batches"
  fix NPE during bookmark creation
  bump agp
  null check axolotl service when getting trust
  check domain name against DNSName to avoid rare crashes
  work around platform bug when getting restrict background
  pulled translations from transifex

* 'peter-dev' of https://gitea.angry.im/PeterCxy/cheogram:
  Call setActive anyway in CONNECTED state
  Add @RequiresApi annotation for dialer integration
  Use StatusHint label to signal connecting state
  set in-call UI active in the connecting state

There is no code that guarantees the state must go through CONNECTING
before being CONNECTED. To avoid potential bugs, call setActive anyway
here.

to stop the linter complaining every second line of code

Currently, if the "CONNECTING" stage of a Jingle call takes a
significant amount of time (say, a second), then there will be a window
where the call is already accepted but the ringing UI is still showing.
This 1) is confusing; 2) allows the user to accept a call twice, which
will cause a crash.

Unfortunately, there is no native equivalent to the "CONNECTING" state.
The only sane choice here seems to be showing the in-call UI even if the
call is still connecting.

This reverts commit 6bd552f6a32ca93826cb491f9b4bd757f9698227.

fixes #4313

This turned out to be a rather unnecessary optimization that might cause
problems with wake locks (the app is no longer awake after the 400ms timeout)

closes #4312

fixes #4211

thank you @singpolyma

Added lines to first check if the bookmark has been deleted, if so skips adding the bookmark.

Previously we would check for the identity at contact presence change,
but the first time the contact comes online (say, during onboarding or
first app use) the contact has never had service discovery done and so
the identity is not known.  Move the identity check to after the service
discovery has certainly been done in order to be sure it works first
time.

While standard TLS encryption is not E2E, it is hardly "unencrypted" and so we
should not be scaring people quite so strongly about it.

fixes #4305

* extended_and_tor:
  Allow extended settings and tor in any build flavour

Trust the setting, don't override it by flavour.

* 'master' of github.com:iNPUTmice/Conversations:
  version bump to 2.10.5 + changelog
  retrieve uncompressed file size in HEAD request
  disable knownFileSize on re-download for pgp encrypted files
  limit posh files to 10k
  ensure downloaded file does not exceed Content-Length reported by HEAD