fix(validators): handle pointer fields in RequiredIf (#334)

[SAY-5]

Fixes #334.

Validation.RequiredIf silently skipped its check whenever either side of the rule was a pointer:

• Destination field is a pointer. When the rule references a *bool/*string field, reflect.ValueOf(dstVal).Kind() is reflect.Pointer. convTypeByBaseKind has no string-to-Pointer conversion, so dstVal == wantVal was never evaluated and the function fell through to the default as True, skip check branch.
• Source field is a pointer to a zero value. val != nil && !IsEmpty(val) is true for *string("") (the pointer is non-nil), so the must be present and not empty contract was not enforced.

Unwrap pointers on both sides:

• A nil pointer on the destination side counts as the field being absent: the rule does not trigger and the dependent field stays optional.
• A non-nil pointer on the destination side is compared against the rule argument by its underlying kind, matching how a non-pointer field of the same type is handled today.
• A new requiredIfValIsPresent helper applies the same nil-and-zero-via-pointer logic to the source field, so *string("") is treated the same as "".

The reproducer from the issue:

type Person struct {
    IsOld      *bool
    Experience *string `validate:"requiredIf:IsOld,true"`
}

Before this PR, both {IsOld: &true, Experience: &""} and {IsOld: &true} reported IsSuccess() == true. With this PR they correctly report false, while {IsOld: &false} and {IsOld: nil} still pass and {IsOld: &true, Experience: &"10y"} still passes.

Tests

• TestIssues_334 in issues_test.go adds five subtests covering each pointer scenario in the issue. Two of them (pointer to empty string is not present, nil pointer is not present) fail on the old code with Result should be False and pass on the new code; the three previously-passing positive cases continue to pass.
• go test ./... -count=1 -race passes locally for the main package and all three locale sub-packages.
• go vet ./... and gofmt -l validators.go issues_test.go are clean.