Fall back to reflection when generated adapter is not found.

[NightlyNexus]

This is handy for development builds where kapt is disabled and models still have @JsonClass(generatedAdapter = true).

Closes #715

[swankjesse]

Not sure about this. Though it’s convenient for development it also hides configuration bugs.

[swankjesse]

Also needs a test to somehow confirm it’s doing the right thing.

What’d you think about enabling this behavior via some Moshi configuration? Perhaps this:

Moshi moshi = new Moshi.Builder()
    .failOnMissingCodegen(false)
    .build()

Or perhaps we just recover gracefully if the reflection adapter is installed:

Moshi moshi = new Moshi.Builder()
    .add(new KotlinJsonAdapterFactory())
    .build()

[NightlyNexus]

Definitely don't want to hide configuration errors. My thought was that development builds would turn off kapt and provide a Moshi instance with the KotlinJsonAdaterFactory is installed, and production builds would turn on kapt and not install the KotlinJsonAdapterFactory. I don't know how common it is to use both in production.

perhaps we just recover gracefully if the reflection adapter is installed

If I understand correctly, that is the behavior of this change?

[swankjesse]

Ahhh, yes I get it. Can you add a test? You’ll need to put it in the reflection adapter’s module so it doesn’t get codegen