Use stage input partitioning partitionCount instead of output partitioning

[gaurav8297]

Description

Instead of using output partitioning scheme's partitionCount, use the input partitionCount for selecting the number of nodes for that stage.

Additional context and related issues

Release notes

(x) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
( ) Release notes are required, with the following suggested text:

# Section
* Fix some things. ({issue}`issuenumber`)

[radek-kondziolka]

Some comments:
0. Very good change with renaming partitioningScheme -> outputPartitioningScheme.

1. I would not add the new field PlanFragment.partitionCount. We can just use outputPartitioningScheme.partitionCount. What is meaning of PlanFragment.partitionCount? I think you mean the number of input partition to the current stage. If yes, it could be taken from children, using StageManager.getChidlren. It is more explicit way to reason about and less confusing. Then we have clear relationship between partitioning: I am HASH partitioned stage and my children's output is HASH partitioned so I can ask them about number of partitions, what do you think?
2. I would fire the rule DeterminePartitionCount if we detect that there is no not-SystemPartitioningHandle. Otherwise, you can have a plan with empty partitionCount and then adaptive hash-partition-count does not work.

[radek-kondziolka]

LGTM. I would add tests when we have a plan with stages:

S0: [any not SystemPartitioningHandle]
S1: [HASH]

and

S0: [HASH]
S1: [any not SystemPartitioningHandle]

[radek-kondziolka]

What's about map

trino/core/trino-main/src/main/java/io/trino/execution/scheduler/PipelinedQueryScheduler.java

Line 847 in e705e1e

Map<PartitioningHandle, NodePartitionMap> partitioningCacheMap = new HashMap<>();

here? Don't we want to key that map with PartitioningHandle and PartitionCount?