Keep former channel_update in channel state

[pm47]

This is very similar to #1918 (ef31de1), but we use the internal
actor state instead of the ChannelData.

Pros:

• conceptually simple, low risk of regression
• ChannelData stays untouched

Cons:

• now there is a var in the channel

Those are the minimal changes to have the simplest diff, but we can
include some improvements made in ef31de1.

[codecov-commenter]

Codecov Report

Merging #1934 (6a1fc84) into master (54fa208) will increase coverage by 0.07%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1934      +/-   ##
==========================================
+ Coverage   87.53%   87.61%   +0.07%     
==========================================
  Files         158      158              
  Lines       12231    12268      +37     
  Branches      505      520      +15     
==========================================
+ Hits        10707    10749      +42     
+ Misses       1524     1519       -5     

Impacted Files	Coverage Δ
...c/main/scala/fr/acinq/eclair/channel/Channel.scala	88.20% <100.00%> (+2.06%)	⬆️
...main/scala/fr/acinq/eclair/db/DbEventHandler.scala	92.53% <100.00%> (-0.43%)	⬇️
...n/scala/fr/acinq/eclair/router/Announcements.scala	100.00% <100.00%> (ø)
...cinq/eclair/blockchain/watchdogs/ExplorerApi.scala	44.89% <0.00%> (-9.39%)	⬇️
...cala/fr/acinq/eclair/payment/relay/NodeRelay.scala	96.06% <0.00%> (-1.58%)	⬇️
...main/scala/fr/acinq/eclair/router/Validation.scala	90.76% <0.00%> (-1.54%)	⬇️
.../acinq/eclair/blockchain/bitcoind/ZmqWatcher.scala	90.00% <0.00%> (ø)
...re/src/main/scala/fr/acinq/eclair/NodeParams.scala	93.26% <0.00%> (+0.03%)	⬆️
...n/scala/fr/acinq/eclair/tor/Socks5Connection.scala	8.66% <0.00%> (+5.32%)	⬆️
... and 1 more

[pm47]

And.... still doesn't work 😒. For the same reasons as #1922 (comment).

If we send multiple CMD_UPDATE_RELAY_FEE in OFFLINE, the previousChannelUpdate_opt will be overwritten.

The root cause of all our problems is that we should not reuse LocalChannelUpdate. The semantics of LocalChannelUpdate revolves around channel availability (connections/disconnections) and trying to minimize events in order to save bandwidth, by deferring some events. It's very different from tracking changes in parameters.