WIP remove lastScope field and deprecated implementations from SimpleObservation

[shakuzen]

This should not be merged until main is 1.18.0 snapshots.

TODO: log warnings, update tests

This is a follow-up to #7255. With removing the lastScopes field on SimpleObservation, we avoid CPU cycles tracking info that is never used and we save memory per SimpleObservation instance.

JOL

We go from 40 bytes per instance to 32 bytes object overhead. See JOL output below. (with CompactObjectHeaders, it gets down to 24 bytes)

io.micrometer.observation.SimpleObservation object internals:
OFF  SZ                            TYPE DESCRIPTION                    VALUE
  0   8                                 (object header: mark)          N/A
  8   4                                 (object header: class)         N/A
 12   4             ObservationRegistry SimpleObservation.registry     N/A
 16   4                         Context SimpleObservation.context      N/A
 20   4           ObservationConvention SimpleObservation.convention   N/A
 24   4       Deque<ObservationHandler> SimpleObservation.handlers     N/A
 28   4   Collection<ObservationFilter> SimpleObservation.filters      N/A
Instance size: 32 bytes
Space losses: 0 bytes internal + 0 bytes external = 0 bytes total

Previous JOL output

io.micrometer.observation.SimpleObservation object internals:
OFF  SZ                            TYPE DESCRIPTION                    VALUE
  0   8                                 (object header: mark)          N/A
  8   4                                 (object header: class)         N/A
 12   4             ObservationRegistry SimpleObservation.registry     N/A
 16   4                         Context SimpleObservation.context      N/A
 20   4           ObservationConvention SimpleObservation.convention   N/A
 24   4       Deque<ObservationHandler> SimpleObservation.handlers     N/A
 28   4   Collection<ObservationFilter> SimpleObservation.filters      N/A
 32   4              Map<Thread, Scope> SimpleObservation.lastScope    N/A
 36   4                                 (object alignment gap)         
Instance size: 40 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total

Benchmarks

We see allocation savings of 72 bytes per operation (explained by the object layout difference above and one less ConcurrentHashMap being instantiated). Note that is irrespective of the number of key-values used, so it will be a higher percent savings for Observations with less key-values.

This PR

ObservationBenchmark.observation                          avgt    2     96.621           ns/op
ObservationBenchmark.observation:gc.alloc.rate.norm       avgt    2    672.000            B/op

1.17.0-M3

ObservationBenchmark.observation                          avgt    2     96.910           ns/op
ObservationBenchmark.observation:gc.alloc.rate.norm       avgt    2    744.000            B/op