fix: Fix incorrect result from casting double to decimal

[zhli1142015]

When converting 0.03456789 to DECIMAL(38, 33), the result generated by Velox is
inconsistent with Spark and Presto. This discrepancy is caused by precision
errors due to long double multiplication. I propose fixing this issue by first
converting the result to the target type(int64_t / int128_t) and then applying rescaling.

Velox

spark.sql("select cast(0.03456789d as decimal(38, 33))").collect
res18: Array[org.apache.spark.sql.Row] = Array([0.034567889999999999999845079187456])

Spark

spark.sql("select cast(0.03456789d as decimal(38, 33))").collect
res16: Array[org.apache.spark.sql.Row] = Array([0.034567890000000000000000000000000])

Presto

presto> select cast(0.03456789 as decimal(38, 33));
                _col0
-------------------------------------
 0.034567890000000000000000000000000
(1 row)

[netlify]

✅ Deploy Preview for meta-velox canceled.

Name	Link
🔨 Latest commit	b64362e
🔍 Latest deploy log	https://app.netlify.com/sites/meta-velox/deploys/67d0c6abc1bf760008e3c569

[zhli1142015]

@rui-mo sn @mbasmanova could you help review this PR? The CI failure is unrelated.

[mbasmanova]

@zhli1142015

The CI failure is unrelated.

Is there an issue about this failure? If not, would you create one and reference it here?

[mbasmanova]

@Yuhta Jimmy, would you help review this fix?

[zhli1142015]

@zhli1142015

The CI failure is unrelated.

Is there an issue about this failure? If not, would you create one and reference it here?

#12603
Thanks.