big data point cloud read

1. Use C instead of C++, JAVA.

2. Chunk read, not row.

3. Analysis own, not  library function, String to value(int, float), ect..

4. 13000 million ascii， At least 4 times faster。