Eigenvector computation in bf_lbo is slow

[pbeckman]

A profiling run is in order. Potential changes include:

• use larger number of desired eigenvectors per UMFPACK call
• implement simple OpenMP parallelism

Time per interval also sometimes grows...

(base) beckman@Pauls-MBP lbo % ./bf_lbo --matPath=../../../../butterfly-LBO-models/sphere_meshes/p/2/ --useOctree --tol=1e-3 --bailIfBottomedOut --freqTreeDepth=3
using FEM matrices loaded from binaries in directory ../../../../butterfly-LBO-models/sphere_meshes/p/2/ (65538 verts)
wrote octree cells to octree_boxes.txt
row tree has depth 5
computed lambda_max = 429820 [6.9s]
building frequency tree with depth 3 (k = 8)
streaming until we bottom out in the space tree
feed: bracket = (-inf, 1.64), num. eigs = 1
- streamed 1 eigs (0.0% of total) in 7.21s
- uncompressed size: 0.500 MB
- total time: 7.21s
- compressed size:   0.500 MB
- compression rate:  1.000
feed: bracket = [1.64, 6.56), num. eigs = 8
- streamed 9 eigs (0.0% of total) in 19.62s
- uncompressed size: 4.500 MB
- total time: 19.62s
- compressed size:   4.501 MB
- compression rate:  1.000
feed: bracket = [6.56, 14.76), num. eigs = 7
- streamed 16 eigs (0.0% of total) in 63.10s
- uncompressed size: 8.000 MB
- total time: 63.10s
- compressed size:   8.001 MB
- compression rate:  1.000

no progress after 20 more min...