[FLINK] Support common stateless scalar expressions

### Description

## Motivation

The velox backend already implements a set of common Flink SQL scalar functions, but the gluten-flink planner has no mapping from Flink function name to velox Substrait operator, so these functions cannot be offloaded to velox today.

Functions in scope:

- Time fields: `YEAR`, `MONTH`, `DAYOFMONTH`, `DAYOFYEAR`, `DAYOFWEEK`, `QUARTER`, `HOUR`, `MINUTE`, `SECOND`, `WEEK`
- String/control: `SUBSTRING`, `COALESCE`, `FROM_BASE64`, `JSON_VALUE`

Additionally, `RexNodeConverter` does not convert TIMESTAMP literals to `TimestampValue.create(epochSeconds, nanos)`, so any literal-TIMESTAMP-bearing expression fails earlier in the conversion path.

## Proposed Capability

The planner should produce a valid Substrait plan for SQL like:

```sql
SELECT
EXTRACT(HOUR FROM ts),
SUBSTRING(name, 1, 3),
COALESCE(name, 'unknown'),
JSON_VALUE(payload, '$.user.id')
FROM ...
```

### Gluten version

main branch

Start by tracing the gluten-flink planner's function-name mapping into Velox Substrait operators, then read the RexNodeConverter timestamp-literal conversion path. Verify coverage for the listed time, string, control, and JSON_VALUE expressions, plus TIMESTAMP literals. Done means the example SQL produces valid Substrait plans and literal-timestamp expressions no longer fail during conversion.

Written by the indexing model from the issue text.