refactor search index builder to store urn parts efficiently (#1937)

[jywadhwani]

Tested the following way

1. Run ./docker/dev.sh with this diff
2. Ingested a new dataset using the following command
   curl 'http://localhost:8080/datasets?action=ingest' -X POST -H 'X-RestLi-Protocol-Version:2.0.0' --data '{"snapshot": {"aspects":[{"com.linkedin.common.Ownership":{"owners":[{"owner":"urn:li:corpuser:fbar","type":"DATAOWNER"}],"lastModified":{"time":0,"actor":"urn:li:corpuser:fbar"}}},{"com.linkedin.dataset.UpstreamLineage":{"upstreams":[{"auditStamp":{"time":0,"actor":"urn:li:corpuser:fbar"},"dataset":"urn:li:dataset:(urn:li:dataPlatform:foo,barUp,PROD)","type":"TRANSFORMED"}]}},{"com.linkedin.common.InstitutionalMemory":{"elements":[{"url":"https://www.linkedin.com","description":"Sample doc","createStamp":{"time":0,"actor":"urn:li:corpuser:fbar"}}]}},{"com.linkedin.schema.SchemaMetadata":{"schemaName":"FooEvent","platform":"urn:li:dataPlatform:foo","version":0,"created":{"time":0,"actor":"urn:li:corpuser:fbar"},"lastModified":{"time":0,"actor":"urn:li:corpuser:fbar"},"hash":"","platformSchema":{"com.linkedin.schema.KafkaSchema":{"documentSchema":"{\"type\":\"record\",\"name\":\"MetadataChangeEvent\",\"namespace\":\"com.linkedin.mxe\",\"doc\":\"Kafka event for proposing a metadata change for an entity.\",\"fields\":[{\"name\":\"auditHeader\",\"type\":{\"type\":\"record\",\"name\":\"KafkaAuditHeader\",\"namespace\":\"com.linkedin.avro2pegasus.events\",\"doc\":\"Header\"}}]}"}},"fields":[{"fieldPath":"foo","description":"Bar","nativeDataType":"string","type":{"type":{"com.linkedin.schema.StringType":{}}}}]}}],"urn":"urn:li:dataset:(urn:li:dataPlatform:foo,mBar,PROD)"}}'
3. Made sure the dataset is both searchable and browsable.

Checklist

• The PR conforms to DataHub's Contributing Guideline (particularly Commit Message Format)
• Links to related issues (if applicable)
• Tests for the changes have been added/updated (if applicable)
• Docs related to the changes have been added/updated (if applicable)

[keremsahin1]

Maybe send out two commits for easier review: one for revert of the revert and the other for the actual fix after revert?

[jplaisted]

Please wait for #1971. You can patch and rebase on top of it if you want to do the work now.

[jywadhwani]

Maybe send out two commits for easier review: one for revert of the revert and the other for the actual fix after revert?

Done. The first commit is revert of the revert (i.e. cherry-pick) and the second is the actual fix.

[jplaisted]

Maybe send out two commits for easier review: one for revert of the revert and the other for the actual fix after revert?

Done. The first commit is revert of the revert (i.e. cherry-pick) and the second is the actual fix.

Neato. That'll make it easier to review :)

[keremsahin1]

Neato

Neato the robot? Didn't get the reference :)

[jplaisted]

Neato

Neato the robot? Didn't get the reference :)

(colloquial, slang) Neat (in the sense of being excellent or desirable).
Oh, neato! My car is fixed.

https://en.wiktionary.org/wiki/neato#:~:text=(file)-,Interjection,Oh%2C%20neato!

Just another way of saying "neat" lol

[keremsahin1]

Just another way of saying "neat" lol

Haha TIL

[jplaisted]

It's easier to write, has more specific asserts, and has better error messages. We did talk about this a few weeks ago :)

[nagarjunakanamarlapudi]

LGTM