[GR-52534] Change digest algorithm and encoding.

[graalvmbot]

Digests are used in multiple places in the image generator, from lambda names to factory method names to symbol names. Shorter digests, and shorter "unique short names", therefore mean smaller image size.

This PR changes the digest from SHA-1 encoded as a hex string (40 bytes) to 128-bit Murmur3 as a Base-62 string (22 bytes). We do not use a standard Base64 encoding because Base64 needs 2 special characters in addition to numbers and letters, and there are no 2 characters that work universally everywhere (Java names, symbol names, ...). We also don't care about encoding speed and never need to decode, so a non-standard encoding does not matter at all.

There is no need for the hash algorithm to be cryptographic. In the worst case, image build can fail because two symbol names are no longer unique. For many usages, there is already a conflict resolution policy in place because the digested input strings themselves are not guaranteed to be unique. A non-cryptographic algorithm also avoids false-positive flagging of the old SHA-1 as a weak cryptographic algorithm without moving to an even longer SHA-2 hash.

In addition, there are a few other tweaks that shorten "unique short names", like capping the length of class names, removing the unnecessary constructor literal for constructors, greatly shortening the deoptimization entry point marker, ...