Sharding should not use hidden indexes

MongoDB 4.4 introduces the possibility to hide an index, this allows to ensure that there are no performance impact on dropping it.
Currently the sharding system does not take into consideration this information and it can use an index even if it's hidden reducing the scope of testing the possibile impact in dropping an index

db.test.insert({a:1})

sh.shardCollection("test.test",{a:1})

-> fail

db.test.createIndex({a:1})

db.test.hideIndex({a:1})

sh.shardCollection("test.test",{a:1})