RangeError when running model with a smaller amount of outputs than in a previous run

[chrispcampbell]

While upgrading En-ROADS to work with the latest published @sdeverywhere/runtime@0.2.4 (which included some refactoring, see #471), I encountered an error in a special case:

• model runs once with normal outputs
• model runs a second time with a smaller subset of outputs/points

In this scenario, the wasm model will reuse a larger outputs buffer than is needed for the second run, so when storeOutputs is called, it fails when copying the array to the internal buffer because the internal view is smaller than the incoming array.

Here's the error (this is from running in NodeJS; in a browser, it may say source array is too long):

Error [RangeError]: offset is out of bounds
    at Float64Array.set (<anonymous>)
    at BufferedRunModelParams.storeOutputs ([worker eval]:3:5106)
    at BaseRunnableModel.runModel ([worker eval]:3:13145)
    at runModel ([worker eval]:3:18439)
    at [worker eval]:3:92

In the refactoring in #471, I had added a lot of tests, but in the one relevant integration test (impl-var-access), it happened to use the same number of outputs when testing both declared and impl variables, so this bug was masked.

I need to fix the storeOutputs method in BufferedRunModelParams to handle the case where the incoming array is larger than the internal view.