[Sparse] Leaf-level bitmask for CPU/CUDA

[yuanming-hu]

Concisely describe the proposed feature

As of v0.5.8 we haven't supported ti.bitmasked(ti.ij, 128).place(x), since we've made an assumption that place's father must be dense. We should remove this assumption.

Additional comments
The ultimate goal of Taichi sparse programming is to make algorithms fully independent of data structures. Assuming the finest level must be a dense node actually harms, since the data mask now has granularity of the dense node dim (e.g. 4x4x8, which is data-structure-dependent) instead of 1x1x1.