ManifestGroup::TaskContext should cache partition spec

[lirui-apache]

Feature Request / Improvement

When we create a TaskContext, the PartitionSpec and Schema are serialized to JSON strings and passed to each BaseContentScanTask. When we later want to inspect schema and spec of the tasks, they need to be deserialized. This can be very expensive if we have lots of columns in the schema. In our use case, the table has over 2k columns.

So I think we should store the schema/spec in TaskContext and pass them to BaseContentScanTask.

Query engine

None

Willingness to contribute

• I can contribute this improvement/feature independently
• I would be willing to contribute this improvement/feature with guidance from the Iceberg community
• I cannot contribute this improvement/feature at this time