Fix spin doctor spurious "missing Wasm file" if app is in another directory

[itowlson]

I ran into this because I was playing with spin doctor using spin doctor -f ~/testing/kvmadness and it reported that my Wasm file was missing. It looks like it currently probes for the file relative to the current directory rather than the app directory.

This fixes that, but it feels a bit verbose, like I am passing more stuff around more places than should be necessary. @lann I'd be grateful for any insight on how you envisaged things like this being structured. I saw that the WasmDiagnostic trait passes the PatientApp that includes the manifest file path, and originally used that to construct the absolute Wasm path in diagnose_wasm.
But it felt like that would end up getting repeated in multiple diagnostics so I moved it into the PatientWasm type. Anyway I don't want to end up spoiling the design on my first outing grin

[lann]

This doesn't seem too bad to me, but another option to consider:

impl PatientWasm {
    // ...

    pub fn local_abs_path(&self) -> Option<PathBuf> {
        match self.source() {
            WasmSource::Local(path) => Some(self.app_dir.join(path)),
            _ => None,
        }
    }
}

Patient/PatientWasm feel like a good place to hang lots of this kind of helper method to ease diagnosis implementation.

[itowlson]

Thanks @lann - I adopted that and it made for a cleaner diff I think. But it did make some of the "when do I use what" a bit fuzzier, so I ended up doing some renames anyway... Anyway ready for review now.

[lann]

Maybe change this to source_path(&self) -> Option<&Path> and drop WasmSource entirely? I think that will be clear enough and we'll need to document these before stabilizing this interface anyway.

[lann]

or maybe local_source_path and local_source_abs_path or something

[rylev]

Does this play well with the changes from #1503?

[itowlson]

@rylev Ugh, good catch. Probably not. It looks like those paths are resolved at the apping stage rather than the raw load stage, and the Wasm doctor wisely relies only on the raw manifest (because the apping stage requires a high degree of validity).

I guess I can bodge around it or rework the loader pipeline to more finely reflect load stages. The latter is my long term goal but I might try the former for now. sigh

[itowlson]

@rylev updated for #1503