rtpsession: Always keep at least one NACK on early RTCP

We recently added code to remove outdate NACK to avoid using bandwidth
for packet that have no chance of arriving on time. Though, this had a
side effect, which is that it was to get an early RTCP packet with no
feedback into it. This was pretty useless but also had a side effect,
which is that the RTX RTT value would never be updated. So we we stared
having late RTX request due to high RTT, we'd never manage to recover.

This fixes the regression by making sure we keep at least one NACK in
this situation. This is really light on the bandwidth and allow for
quick recover after the RTT have spiked higher then the jitterbuffer
capacity.
19 jobs for nacks-fix in 84 minutes and 4 seconds (queued for 2 seconds)
Status Job ID Name Coverage
  Preparation
passed #305914
gst indent

00:00:36

passed #305913
manifest

00:00:26

 
  Build
passed #306005
build fedora x86_64

00:04:33

passed #305983
build nodebug fedora x86_64

00:04:56

failed #305915
build fedora x86_64

00:10:49

failed #305916
build nodebug fedora x86_64

00:10:49

 
  Test
passed #305917
check fedora

00:03:38

passed #305918
integration testsuites fedora

00:06:34

 
  Full Builds
passed #305920
build cerbero android universal

00:20:23

passed #305921
build cerbero cross win32

00:12:40

passed #305922
build cerbero cross win64

00:11:48

passed #305919
build cerbero fedora x86_64

00:14:36

 
  Full Tests
passed #305924
valgrind bad

00:26:19

passed #306102
valgrind ges

00:03:51

passed #306107
valgrind good

00:09:23

failed #305925
valgrind ges

00:18:41

failed #305923
valgrind good

00:29:13

 
  Apps
passed #305926
android universal examples

00:05:52

 
  Deploy
passed #305927
documentation

00:03:29