[Enesuygurs]

e583226 made x-ranges with a number after a wildcard invalid, so 1.x.5, x.1 and x.1.2 now throw. That check only runs in replaceXRange. The same input behind ~ or ^ is still accepted, and the trailing numbers are silently dropped:

new Range('x.1.2')           // throws Invalid comparator
new Range('^x.1.2').range    // '' (same as '*')
validRange('^x.1.2')         // '*'
satisfies('0.0.1', '^x.1.2') // true
new Range('~1.x.5').range    // '>=1.0.0 <2.0.0-0'

^x.1.2 turning into * is the worst case: a mistyped range validates and matches every version.

This PR calls the existing invalidXRangeOrder() check from replaceTilde and replaceCaret as well, so these ranges are rejected the same way the bare x-range is. In loose mode the invalid token is dropped, which matches the existing 1.x.5 || 2.x loose fixture.

Tests: I added rows to test/fixtures/range-parse.js for ~1.x.5, ~>1.*.5, ~x.1, ^1.x.5, ^1.X.5-beta and ^x.1.2. They run under all four option combinations and also feed test/ranges/valid.js. I updated the fixture row counts in test/classes/range.js to match. The full suite passes with 100% coverage, and npm run lint passes.