fix: Don't clone Span in _init_tools

[hoxbro]

Fixes #6386

[TheoMathurin]

Thanks for the lightning fast fix!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.48%. Comparing base (6480c73) to head (26ff7e5).
Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6387   +/-   ##
=======================================
  Coverage   88.48%   88.48%           
=======================================
  Files         323      323           
  Lines       68460    68469    +9     
=======================================
+ Hits        60579    60588    +9     
  Misses       7881     7881           

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

[philippjfr]

This is not safe. We clone models because a user provided Model can only be used within a single plot. What likely needs to happen is that instead of cloning and replacing the models on a per-plot basis we have to search the entire model graph and replace shared instances all at once.

[hoxbro]

Just to be sure, we still clone the CrosshairTool with tool = type(tool)(**properties). This is to avoid cloning overlay=height in the following. The span in overlay=[height, width] is not cloned.

This is an regression and was only reasonly introduced in #6220.

height = Span(dimension="height")
width = Span(dimension="width")
cht1 = CrosshairTool(overlay=[height, width])
cht2 = CrosshairTool(overlay=height)

[philippjfr]

Sorry, misread it, this does look correct.