Investigate switching error grouping key calculation to use something faster

[axw]

We should investigate switching from MD5 to something faster, like xxhash, for calculating error grouping keys. MD5 is cryptographic, which is not necessary for our purposes. xxHash is non-cryptographic, and considerably faster while maintaining high quality hashes.

[graphaelli]

Let's consider adding a feature flag during 7.x for users to opt in to this, based on how big of a change this ends up being.

[axw]

If we move source mapping to an ingest node pipeline (elastic/apm-server#3606), then xxhash may not be a sensible option. We might be able to use murmur3, however this is provided by a plugin and may not be available.

[axw]

Related Elasticsearch issues, proposing new ingest node processor functionality:

• Fingerprinting Ingest Processor elasticsearch#16938
• [WIP] Re-introduce hash processor elasticsearch#47047
• Add "fingerprint" ingest processor elasticsearch#53578

[kruskall]

The implementation now lives at https://github.com/elastic/apm-data/blob/main/model/modelprocessor/groupingkey.go

I'll move this to the apm-data repository