Resource exhaustion test

[podhrmic]

This is an attempt to recreate a simple resource exhaustion test.
The intent is to test whether the received data are dropped once they are processed (i.e. the buffer should never be full unless a single packet is larger than the backing up buffer). Indeed, this test fails and I am not sure why.

The real-life test is to run server.rs and send UDP packets at port 6969 until the buffer is exhausted.

[astro]

Thank you for reporting this issue. Great that I'm not the only one having this problem! It is solved by PR #187.

Please note that your test does not terminate when there is no more problem.

[podhrmic]

Thanks for fixing it. Do you want me to add some termination criteria to the test (assuming we want to keep it)? Plus I would need to fix the CI build.

[whitequark]

So, the underlying cause is that PacketBuffer may use one entry in the metadata ring buffer for the padding, which causes buffer exhaustion when the actual metadata entry is inserted.

I confirm that #187 is the correct way to fix this.

[podhrmic]

Fixed with 6e2da5e