Since we are running web-server and different handling methods will be called from
different threads, it is more appropriate to use concurrent data structures
for exchanging information between threads.
Prior to this change I observed the following exception which is likely to be due `certPaths` not synchronized between threads:
```
Caused by: kotlin.KotlinNullPointerException
at net.corda.node.utilities.registration.RegistrationHandler$reply$1.invoke(NodeRegistrationTest.kt:151) ~[integrationTest/:?]
at net.corda.node.utilities.registration.RegistrationHandler$reply$1.invoke(NodeRegistrationTest.kt:122) ~[integrationTest/:?]
at net.corda.core.internal.InternalUtils.logElapsedTime(InternalUtils.kt:221) ~[corda-core-corda-4.0-snapshot.jar:?]
at net.corda.core.internal.InternalUtils.logElapsedTime(InternalUtils.kt:213) ~[corda-core-corda-4.0-snapshot.jar:?]
at net.corda.node.utilities.registration.RegistrationHandler.reply(NodeRegistrationTest.kt:149) ~[integrationTest/:?]
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) ~[?:1.8.0_161]
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) ~[?:1.8.0_161]
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) ~[?:1.8.0_161]
at java.lang.reflect.Method.invoke(Method.java:498) ~[?:1.8.0_161]
```
Packages re-jig was meant to reduce the size of the jars that are being produced by the
node driver. E.g. previously `MessageState` was in package `net.corda` which resulted in jar file
of around 2MB to be created. Same for `NodeStatePersistenceTests` which was in `net.corda.node`
* ENT-1725: Introduce FlowAsyncOperation, which allows suspending a flow on a custom operation, such as long running I/O requests, notary commit, etc.
* Move async execute to internal, add more tests.
* Add a 30s test timeout
* Update API doc
* Improve logging for NetworkMap requests
* Allow interrupt in polling if the process started successfully
* Put `advertiseNewParameters` back
* Additional log line to indicate when all the nodes are started
* Improve logging and use concurrent map since it is updated from multiple threads
* Change NetworkMap response validity duration and rename parameter accordingly
* Changes following code review from @shamsasari
Transactions are now retried if a communication failure occurs during execute() or commit().
Note that this may result in duplicate entries in the notary request log, and false state
commit conflicts (the notary client would still receive a successful response).
We need to tidy this up in the future.
