Make NAM disablable - bypass dry signal

[marmal95]

This PR implements #431.

Changes:

1. New UI control added - clickable "NEURAL AMP MODELER" logo title.
2. Clicking logo toggles NAM state between enabled/disabled. Title color also changes between white/black - depending on state.
3. Dry signal is passed to output directly when NAM is disabled - meters are still showing audio levels.
4. All UI controls get disabled when whole NAM is disabled.
5. Mouse events on switches are ignored to prevent enabling noise gate or EQ section. Although it would not impact the signal, disabling them make their state consintent with the whole NAM state and make it more readable for user that those signal chain elements are bypassed.
6. Knobs are still editable in disabled state - they do not impact the signal until NAM is enabled again.

"NEURAL AMP MODELER" logo in now clickable. It changes color between white and black - depending if NAM is enabler or disabled. UI controls are disabled when whole NAM was disabled - dry signal is passed to output.

[sdatkinson]

Ok. Thanks for your patience with the long wait.

I want to get this in before #525 so let me see what I can do to shape up anything I need to then get this merged hopefully if nothing big pops up. I'll take care of the unserialization in #526--I'll have more to pile into it thanks to #525 anyways.

[sdatkinson]

Ok. I pulled your branch down to have a look and do some checking.

The punchline is that I don't think that I'm going to accept this PR, unfortunately. But let me explain why.

The core problem is complexity, in short--this "global" switch has a lot of interactions with a lot of things across the whole plug-in. The problem with that is that it opens up the possibility for a lot of bugs due to unforeseen interactions with other parts of the plugin.

For example, when you bypass the plugin by clicking on the title, then reactivate it by clicking a second time, the current implementation enables the output normalization switch. That's a potential problem because if there's a model that's loaded that doesn't have output loudness, then that switch should stay disabled--there are two different inputs to determine the state of that control.

The way to resolve this particular issue itself isn't too hard--if the global bypass is not active and either (A) there's no model loaded or (B) there's a model loaded and it has an output loudness, then set the output normalization switch to non-disabled. I actually started implementing this and refactoring. But then I started running into other problems and stopped since this sort of confirms my worry.

My worry is that this is only one problem, and there might be others. Identifying and verifying that all of them are functioning as intended is this combinatorially-large state space to check. It's not too bad right now, but it would only get worse as more features are added (and adds another thing that needs to be checked every time a feature is added).

What I'm foreseeing is a large amount of time getting sunk into trying to continually debug functionality related to this feature--in addition to this PR, I'm going to be on the hook to maintain it into the future--which is the real deal-breaker. I don't think I can do that.

I get that this feature is potentially useful to some folks of people, so let me try to suggest some alternatives:

• If we re-think #431 to asking for a toggle that works like the IR toggle (i.e. it only affects the model processing, not everything on the plugin) then that can be a manageable feature--I know this because the IR bypass functionality has proven to be well-behaved. Also, one could achieve the designed "bypass" behavior by controlling separately the pieces of this plugin that this touches--EQ, noise gate, model, and IR.
• There are hosts that allow you to disable plugins (e.g. REAPER); consider using one of those. Alternatively, make a feature request to your host of choice to implement it--the other benefit is that this solution would give the functionality once and for all instead of being a chore for every plugin to (re-)implement.
• There are products that play snapshot NAMs--see https://www.neuralampmodeler.com/ for links--some of which may have the functionality you're looking for if I'm not mistaken. If not, you might reach out to them to put in a feature request.
• An approach that I currently use to get a similar effect is that I put an additional "muting" plugin in front of the signal path I want to bypass, then toggle its state. For this, I'd have two paths in parallel--one with the NAM, and one without. An advantage to this is that downstream plugins continue processing (silence), which causes correct tail behavior when bypassing and activating the effect that you won't get with this implementation.
• Lastly, another option is to Fork this repo, merge the feature, and take the responsibility for bugs into your own hands. I definitely don't mind if you choose to do this--part of why I like this being open-source is to give others the ability to customize it to their needs. I even do it myself! 🙂

Hopefully this helps, and again, thanks for putting together this PR. It wasn't obvious to me what the full implications of this feature might be, but now that I've got a little better idea with it in front of me, I do think that it's the right thing to do to pass on it for the repo that I'm maintaining.