Improve telemetry understanding about the results of implicit/explicit restores

### Is your feature request related to a problem? Please describe.

Restore is an action that happens _incredibly_ frequently across our tooling/ecosystem, but we don't have a good handle on how frequently it's actually useful/necessary.

We should be able to say things like "60% of all restores were actually unnecessary", or "implicit restores tend to be unnecessary - users seem to be doing explicit restores beforehand"

### Describe the solution you'd like
We should be able to gather a new piece of MSBuild-based telemetry that fires after the `Restore` target and collects data from that Target + the ambient build to drive this kind of analysis.

Data that we should gather:
* Restore type: implicit (`/restore)` vs explicit (`-t:Restore` or `dotnet restore`)
* Restore scope: workspace (solution/traversal) vs single-project
* measures of the usefulness of the restore (from the NuGet Restore Task outputs):
```xml

```
* the time the restore took - this may be more difficult because this data isn't available inside the MSBuild execution environment

### Additional Context

When adding new telemetry, we must update the in-repo telemetry docs. MSBuild-based telemetry needs to come from the AllowEmptyTelemetry task, but timing information may need to be gathered by the CLI's injected MSBuild telemetry logger.