Assuming the SQL definition below, there are two tables (t1, t2) and one
materialized view (mv1).

When determining whether mv1 is synchronized with its base tables, the
system compares:

Whether mvp1 is synchronized with partition p1 and base table t2

Whether mvp2 is synchronized with partition p2 and base table t2

Optimization in this PR:

The original logic would fetch t2's snapshot information twice. This PR
improves efficiency by caching t2's snapshot in the refresh context to
avoid redundant retrieval.

CREATE TABLE t1
(
    k2 TINYINT,
    k3 INT not null
)
PARTITION BY LIST(`k3`)
(
    PARTITION `p1` VALUES IN ('1'),
    PARTITION `p2` VALUES IN ('2')
)

CREATE TABLE t2
(
    k2 TINYINT,
    k3 INT not null
);

create materialized view mv1
partition by(k3)
as
select * from t1 join t2;

mv1 will has two partition mvp1,mvp2