Performance improvements for formatPath

[jcopi]

The change reduces memory usage of the formatPath function by:

• Using Grow after creating a strings.Builder to reduce the number of future allocations
• Does not collect path segments into a slice using strings.Split, instead if iterates over the segments using strings.Cut so that there is no need to allocate a string header per path segment.
• Changes the interface of Attributes so that the GetAttribute function returns (string, bool) rather than *string. This eliminates the need for a heap allocation of the attribute value in most cases.

The result of this is that formatPath is faster and has a single allocation (assuming the formatted string is <= 2x the length of the pattern).

Running tool: /usr/local/bin/go test -benchmem -run=^$ -bench ^BenchmarkFormatPath$ github.com/IBM-Cloud/go-etcd-rules/rules

goos: darwin
goarch: amd64
pkg: github.com/IBM-Cloud/go-etcd-rules/rules
cpu: Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
BenchmarkFormatPath/3_matching-16         	 7748780	       156.9 ns/op	      96 B/op	       1 allocs/op
BenchmarkFormatPath/5_matching-16         	 4493900	       278.2 ns/op	     112 B/op	       1 allocs/op
BenchmarkFormatPath/10_matching-16        	 2316057	       509.3 ns/op	     208 B/op	       1 allocs/op
BenchmarkFormatPath/3_no_matches-16       	 7766563	       155.1 ns/op	      96 B/op	       1 allocs/op
BenchmarkFormatPath/5_no_matches-16       	 3947868	       302.5 ns/op	     112 B/op	       1 allocs/op
BenchmarkFormatPath/10_no_matches-16      	 2283081	       539.2 ns/op	     208 B/op	       1 allocs/op
PASS
ok  	github.com/IBM-Cloud/go-etcd-rules/rules	10.219s

This also adds benchmarks and tests for the the formatPath function to ensure that functionality has not changed

ENSURE THE FOLLOWING ARE MET:

Below is a brief summary of PR requirements (full list).

• Your PR title summarizes the changes at a high level (not simply "updated X" or "changes Y")
• Your PR description links to the issue/design. If not available, includes a full description for the change.
• Your code contains relevant unit tests.

By opening this PR for review, the author has agreed that these criteria must be met.

By approving this PR, the reviewers have also agreed these criteria have been met and it is ready to be merged.

[jcopi]

Added the original implementation temporarily as a point of comparison in tests and benchmarks.

[kramvan1]

Results are doc'd can remove this from the PR now.

[jcopi]

Benchmark Results

curr_* benchmark runs represent the new implementation in this PR, orig_* runs represent the original implementation.

goos: darwin
goarch: amd64
pkg: github.com/IBM-Cloud/go-etcd-rules/rules
cpu: Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz

Benchmark Name	count	ns/op	B/op	allocs/op
BenchmarkFormatPath/curr_03_matches-16	6861945	171.8 ns/op	112 B/op	1 allocs/op
BenchmarkFormatPath/orig_03_matches-16	2421471	475.7 ns/op	504 B/op	8 allocs/op
BenchmarkFormatPath/curr_05_matches-16	3146415	376.0 ns/op	432 B/op	2 allocs/op
BenchmarkFormatPath/orig_05_matches-16	1477287	821.0 ns/op	984 B/op	11 allocs/op
BenchmarkFormatPath/curr_10_matches-16	1734121	683.4 ns/op	768 B/op	2 allocs/op
BenchmarkFormatPath/orig_10_matches-16	733143	1516 ns/op	1944 B/op	17 allocs/op
BenchmarkFormatPath/curr_50_matches-16	503605	2241 ns/op	4352 B/op	3 allocs/op
BenchmarkFormatPath/orig_50_matches-16	218018	5055 ns/op	8760 B/op	60 allocs/op
BenchmarkFormatPath/curr_03_missing-16	7513495	151.0 ns/op	112 B/op	1 allocs/op
BenchmarkFormatPath/orig_03_missing-16	2917368	436.9 ns/op	280 B/op	8 allocs/op
BenchmarkFormatPath/curr_05_missing-16	3662947	346.7 ns/op	144 B/op	1 allocs/op
BenchmarkFormatPath/orig_05_missing-16	1734278	673.6 ns/op	376 B/op	10 allocs/op
BenchmarkFormatPath/curr_10_missing-16	2029881	571.8 ns/op	256 B/op	1 allocs/op
BenchmarkFormatPath/orig_10_missing-16	936805	1208 ns/op	760 B/op	16 allocs/op
BenchmarkFormatPath/curr_50_missing-16	934234	1165 ns/op	896 B/op	1 allocs/op
BenchmarkFormatPath/orig_50_missing-16	355747	3293 ns/op	2712 B/op	58 allocs/op
BenchmarkFormatPath/curr_all_slashes-16	5011110	223.8 ns/op	64 B/op	1 allocs/op
BenchmarkFormatPath/orig_all_slashes-16	3893077	303.3 ns/op	352 B/op	1 allocs/op
BenchmarkFormatPath/curr_all_patterns-16	2969810	379.1 ns/op	112 B/op	1 allocs/op
BenchmarkFormatPath/orig_all_patterns-16	871940	1222 ns/op	760 B/op	25 allocs/op

[kramvan1]

Cmt should be 2.5x now

[kramvan1]

Can we add a cmt to the GetAttribute that it's been replaced by FindAttribute?

[kramvan1]

@jcopi Thx for all your work here. We would like to get this in now. I added a few last minor updates and then can create the formal PR for finals reviews.