Implement PluginErrorf

[fabrizio-grafana]

What this PR does / why we need it:
This PR implements the PluginErrorf function, which works similarly to DownstreamErrorf.

[CLAassistant]

All committers have signed the CLA.

[ivanahuckova]

Why is this needed? I personally find this a little redundant considering we can do PluginError(fmt.Errorf(format, a...)). So I am wondering why do we need it.

[fabrizio-grafana]

Wouldn't it be the same for DownstreamErrorf then? I'm fine in not having PluginErrorf, but to me it seems more confusing to only have DownstreamErrorf without the plugin error counterpart than having both DownstreamErrorf and PluginErrorf.

[ivanahuckova]

DownstreamErrorf is not really used anywhere and I would rather get rid of that as I personally think it is unnecessary wrapper. But that would be potentially breaking change - so not planning to do that. But also, if we want to add PluginErrorf, I am wondering what is the reason, so we are not unnecessarily adding more code.

[fabrizio-grafana]

I am wondering what is the reason

One of your plugin wraps errors raised by third-party libraries with custom error messages, so both PluginErrorf and DownstreamErrorf can be useful. But I'm fine in either finding a different approach or using PluginError(fmt.Errorf(...))), and similarly for downstream errors.

I would rather get rid of [DownstreamErrorf] [...] But that would be potentially breaking change

Do you think we can at least mark it as deprecated? Proposal here.
I'm fine in not removing it now, since it is a breaking change. But I'm strongly opinionated in favor of marking it as deprecated at least. Or, as I said, it might be confusing to only have DownstreamErrorf without PluginErrorf.

[andresmgot]

My two cents is that you don't need to deprecate de function if we now it's not being used, you can directly delete it (it's an internal function). In any case, if you want to keep DownstreamErrorf and PluginErrorf, I don't see the problem either (having this syntax sugar to avoid PluginError(fmt.Errorf(... if it's a common thing).

[fabrizio-grafana]

it's not being used

Apparently it is used, even though in just a couple of places.

if you want to keep DownstreamErrorf and PluginErrorf, I don't see the problem either

Good! OK maybe I'll wait to merge or close the PR to see if someone else wants to express their opinions, since I'm not sure who should take the final decision here—I can, but I'm not an owner of the library here, so I'm probably not the best person.