Iceberg: Inefficient Equality Delete file handling

[jasonf20]

Currently, the Iceberg plugin will:

1. Load equality delete files in each split. If the same delete file is used by many splits it will be re-read many times
2. Store a map of the deleted rows per loaded delete file. This is inefficient if rows are updated often. It should be enough to store a single map with the row and the max data sequence number in which it was deleted.

[findinpath]

Does #17115 relates to this issue ?

[jasonf20]

@findinpath It addresses 2. but not 1. from what I can tell. PR #18397 addresses both. I have a more detailed explanation here: #18397 (comment)

[ebyhr]

Closing as #21441