[mod_opus] Fix to allow transcoding between different rates

[ticpu]

This is in relation to #2226 and its discussions.

It includes a comprehensive list of patches that were suggested in isolation but that work better as a whole.

Thanks to @celliso1, @shaunjstokes, @magiclin99 and all others who have contributed to fixing this issue.

Obsoletes: #2582, #2623.
Fixes: #2226

[celliso1]

The commit shows redundant use of actual_samples_per_second:

				uint32_t codec_rate = imp->actual_samples_per_second;

				if (!strcasecmp(map->rm_encoding, "opus")) {
					codec_rate = imp->actual_samples_per_second;
				}`

Did you mean this instead ? So only opus is affected:

				uint32_t codec_rate = imp->samples_per_second;

				if (!strcasecmp(map->rm_encoding, "opus")) {
					codec_rate = imp->actual_samples_per_second;
				}`

See #2582 (comment) regarding #2623.

[ticpu]

@celliso1 oh.. my bad. True, it does affect recordings in an unpredictable way like this. I will correct this, thank you.

[ticpu]

Can confirm this did cause issues with G722 not negotiating 8kHz vs 16kHz, we'll be testing that fix soon.

Right now, I also have other issues with our mod_oreka having weird saturated audio and are still diagnosing if it is related to this PR or FreeSWITCH update.

[ticpu]

Moving to draft as hard-coding samples = 480 causes issues in many places other than to the call itself. Media bugs, recordings, captures are all affected.
I'm currently working on a different patch that will use switch_rtp_set_interval to fix it and looks promising.

[ticpu]

Update on this, even my other fix to adjust RTP scaling isn't fully reliable. Some situations still cause either recordings or audio issues. The only way I found to get around all audio issues is to go through the normal transcoding path whenever there's a frequency mismatch.

I'll be testing the patch from @nltd101 in combination with @shaunjstokes one then mark this ready if all parts of the solution end up working. I improved the commit message from PR #2833 to include the details of the fix as well.

I think both commits will be required to have a fully working system using Opus both on the SBC and internally.

Edit: I'll go ahead and test these commits on a couple softphones, then move to the PBX and SBC. We really need Opus to never do near-matches that can result in broken audio. I'll come back in about 2 weeks with the results of our testing.