store: Add parquet module with Arrow schema mapping

Create the foundation for Parquet-based dump/restore: a schema mapping
module that converts relational Table definitions to Arrow Schema objects.

- arrow_schema() maps Table -> Arrow Schema with system columns (vid,
  block tracking, causality_region) followed by data columns
- data_sources_arrow_schema() provides fixed schema for data_sources$ table
- All ColumnType variants mapped (TSVector skipped, Enum -> Utf8)
- List columns wrapped in Arrow List type based on Column.is_list()

Author: lutter

Cargo.lock

@@ -116,6 +116,7 @@ url = "2.5.8"
 ahash = "0.8.11"
 arrow = { version = "=58.0.0" }
 arrow-flight = { version = "=58.0.0", features = ["flight-sql-experimental"] }
+parquet = { version = "=58.0.0" }
 futures = "0.3.31"
 half = "2.7.1"
 indoc = "2.0.7"

Cargo.toml

@@ -38,6 +38,8 @@ pretty_assertions = "1.4.1"
 sqlparser = { workspace = true }
 thiserror = { workspace = true }
 serde_yaml.workspace = true
+arrow = { workspace = true }
+parquet = { workspace = true }
 
 [dev-dependencies]
 clap.workspace = true

store/postgres/Cargo.toml

@@ -24,6 +24,7 @@ mod fork;
 mod functions;
 mod jobs;
 mod notification_listener;
+mod parquet;
 mod pool;
 mod primary;
 pub mod query_store;

store/postgres/src/lib.rs

@@ -0,0 +1,2 @@
+#[allow(dead_code)]
+pub(crate) mod schema;