DOC: explain read_csv/read_table float_precision behavior

[LarytheLord]

Problem

float_precision in read_csv / read_table is currently under-explained: users see option names (high, legacy, round_trip) but not the practical behavior/tradeoffs (issue #64094).

What this PR changes

• clarifies that float_precision controls text-to-float conversion behavior in the C parser
• documents each option explicitly:
  • None / high: default high-precision converter
  • legacy: original lower-precision converter
  • round_trip: converter focused on round-trip-safe text representation
• explains “round-trip” in practical terms (minimizes representation changes when reading then writing values)

Scope

• docs-only change in parser docs shared by read_csv / read_table
• no parser behavior changes

Closes #64094.

Validation

• python3 -m compileall pandas/io/parsers/readers.py
• git diff --check