vlinsert: improve opentelemetry decode perfomance

[vadimalekseev]

Describe Your Changes

TODO:

• Properly encode nested arrays
• Tests

Also added more test cases compared to master to cover all decoding cases.

Also fixed a case that behaved differently compared to the master branch. See this new test case:

VictoriaLogs/app/vlinsert/opentelemetry/opentelemetry_test.go

Lines 315 to 319 in e558503

	// decode bytes of ArrayValue
	data = `[{"scopeLogs":[{"logRecords":[{"timeUnixNano":1234,"body":{"arrayValue":{"values":[{"bytesValue":"Zm9vIGJhcg=="}]}}}]}]}]`
	timestampsExpected = []int64{1234}
	resultsExpected = `{"_msg":"[\"Zm9vIGJhcg==\"]","severity":"Unspecified"}`
	f(data, timestampsExpected, resultsExpected)

- on the master branch it encodes as "[Zm9vIGJhcg==]" which is incorrect because /insert/jsonline encodes this JSON as "[\"Zm9vIGJhcg==\"]"

Fixes #689

                                                         │  /tmp/old.txt  │            /tmp/new.txt             │
                                                         │     sec/op     │   sec/op     vs base                │
ParseProtobufRequest/scopes_1/rows_100/attributes_5-14     19191.5n ±  2%   520.4n ± 2%  -97.29% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_100/attributes_10-14    20428.0n ±  2%   548.9n ± 3%  -97.31% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_1000/attributes_5-14     92.719µ ±  0%   4.511µ ± 3%  -95.13% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_1000/attributes_10-14    92.803µ ±  0%   4.480µ ± 3%  -95.17% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_100/attributes_5-14      35.788µ ± 11%   1.025µ ± 3%  -97.14% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_100/attributes_10-14     36.434µ ±  1%   1.094µ ± 3%  -97.00% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_1000/attributes_5-14    165.066µ ±  0%   8.878µ ± 4%  -94.62% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_1000/attributes_10-14   165.714µ ±  0%   8.917µ ± 3%  -94.62% (p=0.000 n=20)
geomean                                                      57.55µ         2.181µ       -96.21%

                                                         │ /tmp/old.txt  │               /tmp/new.txt               │
                                                         │      B/s      │      B/s        vs base                  │
ParseProtobufRequest/scopes_1/rows_100/attributes_5-14     4.969Mi ±  2%   183.253Mi ± 2%  +3588.20% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_100/attributes_10-14    4.668Mi ±  2%   173.745Mi ± 3%  +3621.86% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_1000/attributes_5-14    10.29Mi ±  0%    211.42Mi ± 3%  +1955.49% (p=0.000 n=20)
ParseProtobufRequest/scopes_1/rows_1000/attributes_10-14   10.28Mi ±  0%    212.86Mi ± 3%  +1971.46% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_100/attributes_5-14     5.326Mi ± 10%   186.119Mi ± 3%  +3394.36% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_100/attributes_10-14    5.236Mi ±  1%   174.408Mi ± 3%  +3231.15% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_1000/attributes_5-14    11.55Mi ±  0%    214.84Mi ± 4%  +1759.51% (p=0.000 n=20)
ParseProtobufRequest/scopes_2/rows_1000/attributes_10-14   11.51Mi ±  0%    213.91Mi ± 3%  +1758.33% (p=0.000 n=20)
geomean                                                    7.410Mi           195.5Mi       +2538.80%

Checklist

The following checks are mandatory:

• My change adheres to VictoriaMetrics contributing guidelines.
• My change adheres to VictoriaMetrics development goals.

[valyala]

@vadimalekseev , thank you for the great optimization of OpenTelemetry data ingestion performance at VictoriaLogs!

[valyala]

FYI, this pull request has been included in VictoriaLogs v1.40.0 release.