Gram matrix computation for more than two tables

[amirsh]

For computing the covariance matrix (crossprod or Gram matrix in the terminology of the VLDB'17 paper), the datasets with two tables work fine. However, for datasets with more than two tables (e.g., Expedia, Flight) a segmentation fault is given as error.
Here is the source code for reproducing:

https://github.com/amirsh/MorpheusPy/blob/90e7174f704e2799faa3074fcdcfe2bcce07959f/examples/covar_expedia.py