Simplify label boundary detection in sample parser

[afrkorsakow]

Optimize line parsing by using str.find() for locating { and str.rfind() for location }

Since metric names cannnot contain { and values are always numeric, escape handling is entirely unnecessary for bounday detection. This imporvemnt scales with line length.

benchmark

git switch master
pytest tests/test_parser.py  --benchmark-only --benchmark-save master
git switch feature/performe
pytest tests/test_parser.py  --benchmark-only --benchmark-compare 0002_master

----------------------------------------------------------------------------------------------------------- benchmark: 2 tests -----------------------------------------------------------------------------------------------------------
Name (time in us)                                                    Min                 Max                Mean             StdDev              Median                IQR            Outliers  OPS (Kops/s)            Rounds  Iterations
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_benchmark_text_string_to_metric_families (NOW)             115.9000 (1.0)      205.4800 (1.0)      128.7905 (1.0)      13.6625 (1.0)      123.5150 (1.0)       9.7862 (1.0)       438;402        7.7645 (1.0)        3587           1
test_benchmark_text_string_to_metric_families (0002_master)     147.6700 (1.27)     319.5870 (1.56)     168.5964 (1.31)     18.2601 (1.34)     161.9480 (1.31)     14.6172 (1.49)      480;383        5.9313 (0.76)       3373           1
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

[csmarchbanks]

Thanks!

With the new UTF-8 support there can be escaped braces in a metric name. I think that means this cannot be done as it could find a set of labels mid-metric and then error. We should probably add a test case for that to make sure however, and it is definitely possible we could continue to optimize the parser performance.