Improve transform.project op to compute default output

Nobody has claimed this yet.

When using the `transform.project` op, it's a common pattern to first create an output with a subset of the dimensions of the input, then projecting along the remaining dimension, e.g. for an input with 3 dimensions:

```groovy
#@ Dataset input
#@ OpService ops

result = ops.run("create.img", [input.dimension(0), input.dimension(1)])
sumOp = ops.op("sum", input)
ops.run("project", result, input, sumOp, 2)
```

It would be great to have the `project` op generate its own output if none is provided (i.e. act as a `Function`), so that calling the op gets as easy as:

```groovy
result = ops.run("project", input, sumOp, dim)
```

Start at the `transform.project` op entry point and inspect how it currently receives and uses the output dataset. Determine how the omitted-output call should derive its dimensions from the input and projection dimension. Done means the op supports both the existing explicit-output form and the requested call that returns a computed result.

Written by the indexing model from the issue text.