Calling `lf.clear()` on an empty `LSHForest` causes a segmentation fault down the line

[cwognum]

When you call clear on an empty LSHForest, it will cause a segmentation fault down the line, e.g. when trying to call layout_from_lsh_forest.

You typically wouldn't have to clear an empty tree, of course, but I had added it in a Notebook where the code cell defining the forest was separate from the code cell populating it with data. I added clear() to the second code cell to prevent myself from accidentally adding the same data twice.

Minimal Reproducible Example

import tmap as tm
import numpy as np

# Initialize a new tree
lf = tm.LSHForest(128)

# Without this line, everything will still work
lf.clear()

# Generate some mock data
data = [tm.VectorUchar(np.random.randint(0, high=2, size=128)) for _ in range(128)]

# Encode the mock data
enc = tm.Minhash(128)
fps = enc.batch_from_binary_array(data)

# Add the data to the tree
lf.batch_add(fps)
lf.index()

# This will now cause a segmentation fault
tm.layout_from_lsh_forest(lf)