Validate multipleOf with exact decimal arithmetic

[jnbdz]

Fixes #106

Motivation

#106 reported multipleOf: 0.001 rejecting values like 1.001, 1.01, 1.02 in the legacy 4.x MultipleOfValidatorFactory. Those cases already pass in the current validator thanks to its epsilon workaround — but the workaround has the opposite defect: the epsilon is absolute (1.1920929e-7, FLT_EPSILON), so any multipleOf smaller than the epsilon accepts every value. For example 0.00000015 validates against multipleOf: 0.0000001 even though it is 1.5× the divisor — a false accept that matters for schemas describing high-precision quantities (8-decimal currency amounts, coordinates).

Changes

Replace the epsilon comparison with an exact remainder computed via BigDecimal on the decimal representation (BigDecimal.valueOf, which the existing toBigDecimal helper already uses). This makes the check exact for the decimal values users write in schemas and instances:

• the Incorrect Vert.x Open API 'multipleOf' validation #106 cases (1.001 / 0.001 etc.) remain valid — binary rounding artifacts no longer need an epsilon;
• non-multiples below the old epsilon are now correctly rejected;
• non-finite values and a zero divisor keep the previous lenient behavior (no error) instead of throwing.

Numbers.remainder had no other callers and is replaced by Numbers.isMultipleOf. The official test-suite multipleOf cases (small numbers, 1e308 overflow with multipleOf: 0.5, float-division-inf) all pass unchanged; full suite green.

Added regression tests covering the #106 values and the sub-epsilon false-accept cases.