Introduce L2CAP infrastructure.

[mickeyl]

L2CAP is the underlying technology powering GATT.
BLE 5 exposes L2CAP COC (Connection Oriented Channels) allowing a streaming API that leads to much higher throughputs than you can achieve with updating GATT characteristics.

The patch follows the established infrastructure very closely. The main components are:

• NimBLEL2CAPChannel, encapsulating an L2CAP COC.
• NimBLEL2CAPServer, encapsulating the L2CAP service.
• Examples/L2CAP, containing a client and a server application.

Apart from these, only minor adjustments to the existing code was necessary.

NOTE: The example programs point out a potential problem in the l2cap stack, probably deep inside Nimble, but we have to investigate that. If you are sending too many data, the stack silently drops it.

NOTE2: I forked the repository over two years ago and unfortunately didn't sync afterwards, so I had to do more merge work than expected to catch up with your changes. I hope everything is correct, but in order to reduce coming diffs, it's probably a good idea to pull this now.

This obsoletes #116

[mickeyl]

Hmm… it fails to compile with ESP-IDF 4.4, probably due to the older C++ standard:

NimBLEL2CAPChannel.h:77:33: error: use of deleted function 'std::atomic<bool>::atomic(const std::atomic<bool>&)'
     std::atomic<bool> stalled = false;
                                 ^~~~~
In file included from /__w/esp-nimble-cpp/esp-nimble-cpp/components/esp-nimble-cpp/src/NimBLEL2CAPChannel.h:14,
                 from /__w/esp-nimble-cpp/esp-nimble-cpp/components/esp-nimble-cpp/src/NimBLEL2CAPChannel.cpp:4:
/opt/esp/tools/riscv32-esp-elf/esp-2021r2-patch5-8.4.0/riscv32-esp-elf/riscv32-esp-elf/include/c++/8.4.0/atomic:71:5: note: declared here
     atomic(const atomic&) = delete;
     ^~~~~~

How are we going to fix this correctly?

[h2zero]

Looks good and thank you for updating this, couple little things to clean up the build.

[h2zero]

Thanks @mickeyl, something weird is happening with this PR, it's not triggering the CI or showing the update. Could you squash these commits and force push? Hopefully that triggers the CI.

[mickeyl]

@h2zero For some versions it still fails ­– this time because of some bogus use of min. I don't think this is inside my code and apparently you're doing some #undef min sorcery in your headers… perhaps I need to replicate that?

[h2zero]

Looks good, just missed one.

[h2zero]

Awesome, thanks @mickeyl

[mickeyl]

@h2zero Excellent, glad that this pull request made it. I can now reduce the diff and concentrate on stabilizing the code and probably exposing the (suspected) bugs in the Nimble l2cap infrastructure. Thanks for maintaining this package, I really appreciate that. Expect further pull requests for the l2cap infrastructure coming over the course of this year.

[h2zero]

Awesome, thank you @mickeyl, looking forward to future PR's 😄