Clustering allows match weight args not just match probability

[RobinL]

This PR allows all clustering functions to take a match weight threshold argument instead of a match probability threshold.

You can now provide either (but not both), which makes the behaviour consistent with e.g. inference.predict()

[RobinL]

Possible chart showing how num clusters varies with mw

import altair as alt
import pandas as pd

cc = cluster_pairwise_predictions_at_multiple_thresholds(
    nodes,
    edges,
    node_id_column_name="my_id",
    db_api=db_api,
    # match_probability_thresholds=thresholds,
    match_weight_thresholds=thresholds_weights,
    output_cluster_summary_stats=True,
)


dc_df = cc.as_duckdbpyrelation().df()

# Define the options for the x-axis
x_axis_options = ['threshold_match_probability', 'threshold_match_weight']

# Create a selection parameter with radio buttons
x_axis_param = alt.param(
    name='x_field',
    bind=alt.binding_radio(options=x_axis_options, name='X-axis: '),
    value='threshold_match_probability'
)

# Base chart with dynamic x-axis based on the parameter
base_chart = (
    alt.Chart(dc_df)
    .transform_fold(
        fold=['threshold_match_probability', 'threshold_match_weight'],
        as_=['variable', 'x_value']
    )
    .transform_filter(
        alt.datum.variable == x_axis_param
    )
    .add_params(x_axis_param)
    .encode(
        x=alt.X('x_value:Q', title='X-axis')
    )
    .properties(width=400, height=150)
)

# Define the subcharts
num_clusters = (
    base_chart.mark_line()
    .encode(
        y=alt.Y("num_clusters:Q", title="Number of Clusters")
    )
    .properties(title="Number of Clusters vs X-axis")
)

max_cluster_size = (
    base_chart.mark_line()
    .encode(
        y=alt.Y("max_cluster_size:Q", title="Max Cluster Size")
    )
    .properties(title="Maximum Cluster Size vs X-axis")
)

avg_cluster_size = (
    base_chart.mark_line()
    .encode(
        y=alt.Y("avg_cluster_size:Q", title="Average Cluster Size")
    )
    .properties(title="Average Cluster Size vs X-axis")
)

# Combine the charts
combined_chart = alt.vconcat(
    num_clusters, max_cluster_size, avg_cluster_size
).resolve_scale(y="independent")