perf: Add metric for time spent casting in native scan

[andygrove]

Which issue does this PR close?

N/A

Rationale for this change

Make it easy to see how much of the native ScanExec time is spent casting columns to different types (this usually means unpacking dictionaries).

Example from TPC-DS q9:

DataFusion metrics in native explain output:

metrics=[
  output_rows=2097152, 
  elapsed_compute=21.847892ms, 
  cast_time=21.631731ms]

Full plan:

AggregateExec: mode=Partial, gby=[], aggr=[count, avg, avg], metrics=[output_rows=1, elapsed_compute=9.481194ms]
  ProjectionExec: expr=[col_1@1 as col_0, col_2@2 as col_1], metrics=[output_rows=400519, elapsed_compute=50.596µs]
    FilterExec: col_0@0 IS NOT NULL AND col_0@0 >= 81 AND col_0@0 <= 100, metrics=[output_rows=400519, elapsed_compute=4.753725ms]
      ScanExec: source=[CometScan parquet  (unknown)], schema=[col_0: Int32, col_1: Decimal128(7, 2), col_2: Decimal128(7, 2)], metrics=[output_rows=2097152, elapsed_compute=21.847892ms, cast_time=21.631731ms]

What changes are included in this PR?

How are these changes tested?

[andygrove]

@viirya @comphead I have addressed feedback. Thanks.