reject negative requested_digits in ToFixed

[Ramya-9353]

Repro: ToFixed(1e-30, -1000, &builder) on a buffer sized by the header's own formula (1 + kMaxFixedDigitsBeforePoint + 1 + kMaxFixedDigitsAfterPoint + 1, so 163 bytes) returns true and writes 1000 characters; ASan reports a heap-buffer-overflow write in CreateDecimalRepresentation, and the length written is exactly -requested_digits.

Cause: requested_digits is bounded only from above, so a negative count reaches CreateDecimalRepresentation as decimal_point == -requested_digits and AddPadding emits that many characters. The lower bound existed only as the DOUBLE_CONVERSION_ASSERT in DoubleToAscii, which is compiled out under NDEBUG. ToFixed(1e-30, INT_MIN) additionally trips UBSan on the negation in fixed-dtoa.cc:389.

Fix: reject negative requested_digits up front, matching the guards ToExponential and ToPrecision already carry. Values 0 to kMaxFixedDigitsAfterPoint are unaffected.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[floitsch]

This is arguably something the caller should check and not do, but I'm in favor of catching it ourselves.
The patch is small enough that I will accept it without you signing the CLA. (That said: I recommend signing it, as it makes it easier to contribute to Google projects).
Thanks.

[Ramya-9353]

Appreciate the merge, and thanks for taking it even without the CLA signed.