Inconsistent curve basis handling about hdblackbird

[dedoardo]

From a read through the code I noticed the following issues:

• We currently rely on the cycles tokens to distinguish whether a curve is a cylinder or flat. The usd specification states that it is something that should be deduced from the presence of widths or not. We should have both options.
• The wrap mode of curves seems to be ignored (same goes for basis/type).
• I read that for cubic curves to linearly interpolate a primvar you use 'Varying' but in the code I see 'Vertex'
• I miss the context for the older curves (closing cylinders as tubes?). I wonder if we still need it...

In general I think the code could use being revisited, possibly with a simple test scene containing all possible combinations of inputs.