Misc P2P improvements

[fcecin]

This PR in a nutshell:

• Consolidated 3 P2P thread pools into just one; this means Message object is handled in the same net thread that runs the P2P engine, instead of being offloaded to a separate "incoming Message processing" thread pool (the old ManagerBase::threadPool_). If there is need to offload particular tasks from net threads, this is done in the Message handler itself, and whatever higher-level thread pool(s) do this, they are of no concern to the P2P/Net component
• Removed distinction between Client and Server sessions/connections at the P2P Manager level
• Hardened async networking code in the P2P engine (ManagerBase, Session)
• Simplified socket lifetime (immediately close sockets)
• Cleaned up code
• Fixed bugs
• Improved mutex use
• Improved logging messages

Detailed:

• removed posting of Message objects in a separate thread pool for processing; now handleMessage() is done in the same thread that pumps the bytes out of the peer socket
• single io_context in P2P::ManagerBase, removed io_context from P2P::Client and P2P::Server/P2P::ServerListener
• single ASIO thread pool in P2P::ManagerBase, removed threading from P2P::Client and P2P::Server
• Removed Client, Server, ServerListener classes, and moved all networking functionality into a ManagerBase::Net helper class
• Fix bug in Server socket acceptance where an error in accepting a connection results in the server never accepting a connection again
• Fix Session::do_close()
• Remove SO_REUSEADDR from P2P TCP listen socket
• Add SO_LINGER(0) option to all sockets and removed socket.shutdown() from Session::do_close()
• Add XTRACE and FATAL log levels
• Add dynamic P2P listen port number allocation to relevant networked tests (mostly the ones that have been failing randomly with socket disconnections: state, rdpos, p2p; and some others)
• Changed logical location logging to append the serverPort_ of the P2P node after the instance ID# (now logs looks like " #nn:nnnn ")
• Improved several P2P logging messages
• Misc refactors/fixes

NOTE: I reviewed Sonarqube's test coverage complaints and I think there's nothing relevant there.

[fcecin]

Set this branch to DRAFT again (WIP) because I will be adding more features to it.

Namely:

• Unit test nodes will all use unique ports
• Improve logging with more levels (FATAL & XTRACE)
• Fix clargs.h vs. Options (take options file as parameter and nothing else, move the rest to Options)

I will also try to write some test coverage to obtain that satisfying green checkmark from Sonarqube.

EDIT: disabling command-line args is not obvious, because the Options "singleton" does not look like a singleton in the process sense, but in the blockchain object sense. in unit testing you can have multiple Options objects, so you still need to be able to load and set process-wide options that apply to everything. For example, we don't want custom log levels or threading parameters for each simulated blockchain in unit testing.

[bdk-cpp-sonarqube-app[bot]]

Quality Gate failed

Failed conditions
38.4% Coverage on New Code (required ≥ 80%)

See analysis details on SonarQube

[fcecin]

Unfortunately:

• Code coverage % actually goes down the more you change other files and Sonarqube considers it "new code"
• Covering exception handlers is a waste of time; see ManagerNormal::handleMessage(): the default: label is "uncovered" because we generate an exception inside of Message::type() that is invoked as the case statement. Removing the defensive default: label just to satisfy Sonarqube is arguably a bad idea.

I don't think we need to spend time refining code coverage of things like exception handlers. Most of them are there preventively.