Tags: will0815/pinot

Enhance IndexingConfig to have SegmentPartitionConfig.

The current indexing config has a map of column to partition function
name, which is not sufficient information. Also, it imposes the burden
on client of IndexingConfig to parse this info and build
SegmentPartitionConfig. This change embeds SegmentPartitionConfig inside
of IndexingConfig, thus solving both the problems.

1. Added partition config inside of indexing config.
2. Separated out number of partitions as a first class member of
partition config, as opposed to embeded in partition function string.
3. Moved the config classes to config, as IndexingConfig resides there.
4. Enhanced unit tests to cover the changes.

Fix to recompute routing table when instance config changes in Helix (a…

…pache#1260)

Fix to recompute routing table when instance config changes in Helix

Change email subject to Thirdeye Alert (apache#1251)

Change email subject to Thirdeye Alert (apache#1251)

Fix the rest API for getting table/segment metadata (apache#1250)

Add backward compatible for API:
1. /tables/{tableName}/segments/{segmentName}
2. /tables/{tableName}/segments

Fix the NPE for realtime segment (apache#1238)

In realtime consuming segment, we first add value to dictionary then update inverted index.
If we query a column with inverted index, it is possible that the inverted index has not been generated.
The fix is filter out all null inverted index.
This will not affect the accuracy of the results because we just ignore a single under-indexing record.

Log an info instead of an error for data schema mismatch (apache#1236)

Remove keyword 'LIKE' from PQL grammar.

Removing the keyword 'LIKE' from PQL grammar as it is not supported
currently.

Configuration support for token bucket scheduler

Make token lifetime and number of tokens configurable. Changed
default for lifetime to 100ms and number of tokens dependent on
total number of threads

[TE] Add grouping stage after anomaly detection stage (apache#1226)