Changes proposed in this pull request:

Change ProductQueryExtended join

Currently, using innerJoin() only returns products that have a record in the purchase_counts table. As a result, it could render the paddingQuery useless, especially in cases such as new shops where a majority of products may not have any purchase records. Using leftJoin() resolves this issue.

Exclude purchase pattern IDs from paddingQuery

Currently product IDs queried from the purchase_patterns table are not excluded from the paddingQuery. This can result in returning a query with fewer products that the designated limit when a duplicate ID is returned by the paddingQuery. The _mergeIds() method examines any existing id property of the paddingQuery and excludes any of the product IDs found in purchase_patterns before executing the query.

Add ability to filter IDs queried from purchase_patterns

This was something I tried to resolve in a previous pull request but was not merged for the reasons outlined in this comment. While I initially thought I could resolve the issue using method suggested in the comment, I found that there are scenarios (especially when using a limit) where it does not work as desired.

{% set customersAlsoBought = craft.purchasePatterns.related(
    product,
    3,
    craft.products.relatedTo(product.myCategory).availableForPurchase(1)
).availableForPurchase(1).all() %}

Example 1:
The related() above returns three products, but they all have availableForPurchase set to false. They are subsequently filtered by .availableForPurchase(1), returning an empty array.

Example 2:
The related() above retrieves 1 product from purchase_patterns, filling in the remaining two products with the paddingQuery. The two products have availableForPurchase set to true, but the product from purchase_patterns does not. The query is subsequently filtered by .availableForPurchase(1), returning an array of two products instead of three.

I added a filters parameter and deferred limiting until a second query that applies the filters to product Ids from purchase_patterns. Doing so allows these relationships to be queried with criteria while still ensuring results that consistently match the limit. The query above would be revised to:

{% set customersAlsoBought = craft.purchasePatterns.related(
    product,
    3,
    craft.products.relatedTo(product.myCategory).availableForPurchase(1),
    { availableForPurchase: 1 }
).all() %}

This could also be useful in other situations, such as only retrieving purchase patterns for certain product types (e.g. excluding replacement parts so only retail products are returned).

Thanks for your consideration!