refactor: improve speed of BottomUp

[probberechts]

This PR makes some small modifications to improve the speed of the BottomUp segmentation algorithm, especially when applied to long time series.

---

Below is a simple benchmark:

import time
import ruptures as rpt

n_samples, n_dims, sigma = 100_000, 1, 2
n_bkps = 100  # number of breakpoints
signal, bkps = rpt.pw_constant(n_samples, n_dims, n_bkps, noise_std=sigma)

t1 = time.perf_counter(), time.process_time()
algo = rpt.BottomUp(model="normal").fit(signal)
my_bkps = algo.predict(n_bkps=100)
t2 = time.perf_counter(), time.process_time()
print(f" Real time: {t2[0] - t1[0]:.2f} seconds")
print(f" CPU time: {t2[1] - t1[1]:.2f} seconds")

Current implementation: 46.08 seconds
Adapted implementation: 6.94 seconds

[oboulant]

@probberechts Thx for the great work !
Can you push some fake changes : indeed it is strange, your created PR did not trigger the usual automated checks (tests and everything).

[oboulant]

This is an example of a recent PR for which I could allow the usual checks to run for a first time contributor : #315

[probberechts]

I don't know why it didn't work, but I think you should be able to run the workflows now.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (03ff932) 98.77% compared to head (0347762) 98.77%.
Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #309   +/-   ##
=======================================
  Coverage   98.77%   98.77%           
=======================================
  Files          40       40           
  Lines         978      980    +2     
=======================================
+ Hits          966      968    +2     
  Misses         12       12           

Flag	Coverage Δ
unittests	98.77% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[oboulant]

On my end :
Current implementation: 17.23 seconds
Adapted implementation: 1.36 seconds

🚀

[oboulant]

LGTM