Fall back to floats for out-of-range decimal literals

[osipovartem]

Summary

• keep planning representable numeric literals as Decimal when parse_float_as_decimal is enabled
• fall back to Float64 when a fractional or scientific literal exceeds Arrow Decimal precision/scale limits
• preserve the existing error for out-of-range integer literals, avoiding silent precision loss for integer syntax
• share the existing float parsing/error path between normal and fallback planning

Motivation

Snowflake accepts values such as 1.7976931348623157e+308 and 2.2250738585072014e-308 as DOUBLE. With decimal-literal parsing enabled, DataFusion previously rejected them before target-type coercion because their Decimal scale or precision cannot be represented by Arrow.

This prevented otherwise valid inserts into FLOAT columns. The fallback is limited to syntax that already denotes a fractional/scientific numeric value and only runs after Decimal planning reports an unsupported representation.

Validation

• cargo +1.97.0 test -p datafusion-sql --test sql_integration parse_decimal
• cargo +1.97.0 test -p datafusion-sql test_parse_decimal
• cargo +1.97.0 clippy -p datafusion-sql --all-targets -- -D warnings
• cargo +1.97.0 fmt --all --check
• downstream Rustice data-types/float.slt: 18/19 -> 19/19