[MAINT] `linker._intermediate_table_cache` logs multiple entries when creating term frequency tables about splink HOT 3 CLOSED

@RobinL, I've quickly drafted up a potential solution to address this issue, available for review here.

The proposed changes include:

1. Removal of Templated Names from the Cache: For example, the name __splink__df_concat_with_tf will no longer appear in the cache.

2. Addition of a Method for Templated Name Checks: This method sequentially checks a list of provided templated names. This means that if you wish to search for both tf_first_name and concat_with_tf; __splink_df_tf_first_name is return immediately if found; otherwise we search for __splink_df_concat_with_tf and return it if found. This functionality is only intended for use with term frequency tables to determine if they have been generated without requiring any SQL hashing.

3. Term Frequency Table Registration Changes - Given the changes above, table registration will now overwrite the originally registered tf table. i.e. if __splink__df_concat_with_tf already exists (with any given hashed name), we will now overwrite this table with the table being registered.

It's important to note that this update is based on the assumption (which needs verification and may be incorrect) that in any single Splink session, only one term frequency table should be assigned per column or input nodes table.

Consequently, instead of generating two tables when running compute_tf_table or initialise_df_concat_with_tf, a single table will now be generated and its corresponding templated name will be utilised for cache references.

from splink.duckdb.duckdb_linker import DuckDBLinker
import pandas as pd

from tests.basic_settings import get_settings_dict

df = pd.read_csv("./tests/datasets/fake_1000_from_splink_demos.csv")

linker = DuckDBLinker(
    df,
    get_settings_dict(),
)

# Generates our primary tf table for 'first_name'
tf_table = linker.compute_tf_table("first_name").as_pandas_dataframe()

linker.register_term_frequency_lookup(tf_table, "first_name")  # errors
linker.register_term_frequency_lookup(tf_table, "first_name", overwrite=True)  # registers

from splink.

Link to our discussion about this so it doesn't get lost:
https://asdslack.slack.com/archives/D02TEUPB6H3/p1700482818591709

from splink.

Closing as this is stale and we are content with the current caching methodology.

from splink.