Revisit sparse route

[danielballan]

Current we overload the /array/* routes for sparse data, but I wonder if we should use /sparse/*.

The sparse data supports a different set of formats. That alone isn’t an iron-clad reason to make a separate route, as different arrays support different formats already, depending on dimensionality and metadata. But the difference here is greater.

The sparse route may also grow other parameters in the future that do not make sense of (dense) arrays. And vice versa. For example, we might potentially add a query parameter to convert between different sparse representations.

[danielballan]

Kicking this to v2. Not 100% sure what the right move is, so we'll leave it alone for now.