Update/dash aix360 heart

[budescode]

Issue for app: none (dependency update, no linked issue)

App pull request

• This is a new app
• I am improving an existing app (redesigns/code "makeovers")

About

Brings dash-aix360-heart to current package versions and makes it
installable with uv or pip from a fresh clone.

• Declare the nine direct dependencies in pyproject.toml and lock with uv
• Regenerate requirements.txt from uv.lock so pip users still work
• Bump aix360 0.2.1 -> 0.3.0
• Bump scikit-learn 0.24.2 -> 1.3.2, capped below 1.4 because aix360 uses an argument scikit-learn 1.4 removed
• Bump numpy 1.20.3 -> 1.26.4, capped below 2 for scikit-learn 1.3
• Bump pandas 1.2.4 -> 3.0.5
• Bump cvxpy 1.1.13 -> 1.7.5
• Bump dash 1.20.0 -> 4.4.1
• Bump dash-bootstrap-components 0.12.2 -> 2.0.4
• Bump plotly 4.14.3 -> 7.0.0
• Bump gunicorn 20.1.0 -> 26.2.0
• Drop tensorflow, torch, keras, Django, xgboost, shap, lime and the other packages the app never uses; requirements.txt goes from 96 lines to 54
• Drop dash-core-components, dash-html-components, dash-renderer and dash-table, merged into dash
• Require Python 3.11 or 3.12; raise runtime.txt from 3.7.6 to 3.12.11
• Replace dash_core_components and dash_html_components imports with from dash import dcc, html
• Replace app.run_server with app.run
• Replace dbc.FormGroup, removed in dash-bootstrap-components 1.0, with html.Div
• Replace Series.append with pd.concat and Series.iteritems with Series.items in utils.py, both removed in pandas 2.0
• Update the Python version in the README setup instructions

The scikit-learn and numpy caps are the newest versions that work with
the newest aix360. aix360's FeatureBinarizer still calls
OneHotEncoder(sparse=False), removed in scikit-learn 1.4, on both the
0.3.0 release and current master. Lifting the caps only needs an aix360
release with that call updated.

Verified on Python 3.12: the data loads, the model trains, the
explanation table and GAM data compute, and the callback produces
figures for every combination of the two dropdowns.

• Playground deployment URL (new version):
• Current gallery app URL: (delete this line if inapplicable)

Workflow

• I have created a branch in the appropriate monorepo, and the
  elements necessary for successful deployment are in place.
• If the app is a redesigned and/or restyled version of an
  existing gallery app, I've summarized the changes requested in the
  appropriate Streambed issue and confirm that they have been applied.
• If the app is on the Dash Gallery portal, I have added a link to
  the GitHub repository for the source code in the portal description.
• If the app is a reimplementation of a Python gallery app for the
  DashR gallery, the app in this PR mimics, as closely as possible,
  the style and functionality of the existing app.
• I have removed all Google Analytics code from the app's
  assets/ folder.

The pre-review review

I have addressed all of the following questions:

• Does everything in my code serve some purpose? (I have removed
  any dead and/or irrelevant code.)
• Does everything in my code have a clear purpose? (My code is
  readable and, where it isn't, it has been commented appropriately.)
• Am I reinventing the wheel? (I have used appropriate packages to
  lessen the volume of code that needs to be maintained.)

Post PR (at merge time)

• When you are merging, make sure to write one of the following tags in the commit message (or it will default to patch):
  • #patch - An app has been updated or fixed
  • #minor - A new app has been added, or an app has been significantly reworked
  • #major - Breaking changes, make sure to discuss with dash-core before using this tag