Batch text and line annotation rendering
This commit makes LineRenderer and TextRenderer to batch line and
text rendering. Batching them has shown a significant speedup,
especially when using CUDA.
Results from testing on a developer machine, rendering on a 1024x1024 canvas:
| Before | After | Speedup | |
|---|---|---|---|
| Serial | 41 | 41 | 1.00x |
| OpenMP | 40 | 34 | 1.17x |
| CUDA | 650 | 120 | 5.41x |