Core: Allow reading metadata table when scanning table with dropped partition source field

[gabeiglio]

To replicate this issue run the following spark queries:

CREATE TABLE ns.barfoo (dateint INT,hour INT, data STRING) USING iceberg PARTITIONED BY (dateint, hour);
INSERT INTO ns.barfoo (dateint, hour, data) VALUES (20250101, 1, 'a'), (20250102, 1, 'a'), (20250102, 2, 'a');
ALTER TABLE ns.barfoo DROP PARTITION FIELD dateint;
ALTER TABLE ns.barfoo DROP COLUMN dateint;
SELECT * FROM ns.barfoo.files;

[huaxingao]

Please add a short comment to explain why we intentionally bypass the compatibility check here

[Fokko]

I agree with @huaxingao that a comment would be nice there, apart from that, LGTM! 👍

[gabeiglio]

@huaxingao @Fokko thanks for the review! Added a comment 👍🏼

[singhpk234]

Thank you for the fix @gabeiglio ! LGTM as well

[singhpk234]

Thank you for the fix @gabeiglio ! Thank you for the reviews @Fokko @huaxingao @bryanck !