encodeCbor() throws on objects with non-ASCII keys

encodeCbor() under-allocates its buffer when an object key contains multi-byte UTF-8 characters, then throws while writing:

import { encodeCbor } from "@std/cbor";

encodeCbor({ "é☃é☃é☃é☃": 1 });
// RangeError: offset is out of bounds
The cause is in calcEncodingSize(): object keys are sized as y.length bytes (cbor/_common_encode.ts:58), but encodeString() writes UTF-8, which needs up to 3 bytes per UTF-16 code unit. String values already account for this with x.length * 3 on line 34. Keys don't.

The fix looks like a one-liner: size keys the same way the value branch does. Happy to PR it.

Spotted by @bartlomieju while reviewing #7250.

Start in cbor/_common_encode.ts at calcEncodingSize() line 58 and compare object-key sizing with the string-value branch around line 34. Run the encodeCbor example with the non-ASCII key; done means it no longer throws while writing and successfully encodes the object.

Written by the indexing model from the issue text.