Add minimal support for ssl module using Mbed TLS and sockets

[pguyot]

Details of changes:

• Add support for ssl client in binary and passive modes, with no certificate verification
• Add APIs to otp_socket so it can be called from ssl bio callbacks
• Fix a bug in lwIP otp_socket's recv revealed by ssl tests
• Fix a bug in BSD otp_socket's recvfrom revealed by refactoring
• Fix a bug in esp32 tests where main context and its resources were not properly destroyed
• Update documentation and workflows to reflect the requirement on Mbed TLS
• Fix exported types of inet module

This code was tested on:

• Pico-W
• ESP32 using ESP-IDF 5.1 release branch
• Unix (macOS)

using atomvm_netbench associated test.

The test takes 0.5s with Erlang/OTP or AtomVM on macOS.
It takes 7.3s then 1.0s on ESP32
It takes 12.0s then from 2.2s to 2.6s on Pico-W

These changes are made under both the "Apache 2.0" and the "GNU Lesser General
Public License 2.1 or later" license terms (dual license).

SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later

[bettio]

Looks all good. Maybe this might be the first component implementing this structure: #930
Otherwise we can just merge it, and do this on a second step.

[fadushin]

Is it your feeling that this is not mature enough to have a place in the Programmer's Guide? Do we want to implement PXIX certificate chain validation first? (As it is, I believe the current implementation would be vulnerable to MitM attacks, without authentication).

[bettio]

Looks all good. Maybe this might be the first component implementing this structure: #930 Otherwise we can just merge it, and do this on a second step.

There are still a lot of open points in #930 proposal, so let's just focus in merging this PR, and let's take care of reorganizing code as soon all open points are closed.

[bettio]

Is it your feeling that this is not mature enough to have a place in the Programmer's Guide? Do we want to implement PXIX certificate chain validation first? (As it is, I believe the current implementation would be vulnerable to MitM attacks, without authentication).

Proper documentation for an ssl module would both require a stable API and a reasonable amount of work, in order to explain caveats and how it differs from OTP one. This kind of work deserves a standalone PR, so let's merge this PR and document ssl module properly as soon as we are ready.