Improved receiver buffer implementation

[maxsharabayko]

This PR introduces a new receiver buffer.
To start with, both old and new versions of the receiver buffer are available.
A build option ENABLE_NEW_RCVBUFFER can be used to build SRT with the new version.

Main Features

• Lower latency: not waiting for packets to be acknowledged to allow reading. Resolves the minimum 10ms latency of SRT.
• Handling of sender's drop requests. The old version does not allow to properly drop a message in the middle of the receiver buffer, limiting live CC.
• Out of order reading. SRT with disabled TSBPD handled the "order" flag O of a data packet, allowing to read packets out of order. It is not working properly with the old receiver buffer.
• TBD

Addresses #1674

Description to be improved.

TODO

• In file config currently ACK and isRcvDataReady are not handled properly. See also Transmission.FileUpload test.
• Fix the build with -DENABLE_NEW_RCVBUFFER=OFF.
• Rebase on the latest SRT master.
• CRcvBufferNew::getRcvDataSize(..): counting acknowledged packets requires modifications.
• Check receiver buffer fullness MAvg (takes 2 seconds, which is too long). The same behavior is in the old RCV buffer. Create a separate issue?

[quink-black]

1. utilities.h is included by common.h
2. common.h use something defined by utilities.h, e.g., HAVE_FULL_CXX11
3. Something defined by common.h can be useful for utilities.h, e.g., SRT_ASSERT, SRT_STATIC_ASSERT, but can't due to circular header include

utilities.h is more than utility, it needs some cleanup work. Before that, I prefer the "include what you use" method, and use a separate header file. Of course, the cleanup work can be done after the patch. So the comments above doesn't related to the patch strictly.

[gou4shi1]

I finally found time to look at this great refactor.
It can simply #2046 a lot :)
However, it seems to lack the optimization introduced by #2026

[maxsharabayko]

However, it seems to lack the optimization introduced by #2026

Yes, this optimization is missing.
The ultimate goal is to eventually share a single receiver buffer between all member sockets.
However, it is not there yet. But the optimization is useful. Maybe it makes sense to drop all packets up to the group base before reading to minimize the scope of the reading function. 🤔