fix(rsa): reject zero-modulus JWK keys and throw on BigInteger divide-by-zero

[Kr0emer]

Summary

This PR fixes an RSA robustness/security issue where a malformed JWK RSA public key with n = "AA" (decoded modulus = 0) could be accepted during import.
Combined with BigInteger mod(0) behavior, RSA public operations could silently degrade to deterministic wrong output.

Root Cause

• RSASetPublic accepted non-empty modulus/exponent strings without validating parsed numeric values.
• bnpDivRemTo returned silently when divisor was zero, so modPowInt(e, 0) did not fail loudly.

Changes

• ext/rsa.js
  • Hardened RSASetPublic with numeric checks after parsing:
    • reject n <= 1
    • reject invalid/non-positive e
• ext/jsbn.js
  • bnpDivRemTo now throws "BigInteger divide by zero" when divisor is zero.
• test/qunit-do-crypto.html
  • Added regression tests:
    • RSASetPublic rejects zero modulus
    • KEYUTIL.getKey rejects JWK RSA with n=0 (n: "AA")
    • BigInteger.modPowInt(..., BigInteger.ZERO) throws

Verification

• Reproduced before fix:
  • zero-modulus JWK was accepted
  • modPowInt(e, 0) returned 0
• After fix:
  • zero-modulus JWK is rejected with Invalid RSA public key
  • modPowInt(e, 0) throws BigInteger divide by zero
• Ran tests:
  • npm test (under npm/) passes (6 passing)