Fix ExponentialBucketTimeRange: Unsafe calculation, unpredictable bucket count, and floating point Prometheus issue

[AkramBitar]

Title
Fix ExponentialBucketTimeRange: Unsafe calculation, unpredictable bucket count, and floating point Prometheus output

Description
Problem
The ExponentialBucketTimeRange function in platform/common/utils/metrics.go has multiple critical issues causing memory waste and poor Prometheus output quality:

1. Unsafe Calculation Method
   Uses math.Exp(math.Log(...)) which compounds floating-point errors and is mathematically equivalent to the simpler math.Pow().

2. Unpredictable Bucket Count
   Loop condition for v <= interval produces 9-11 buckets when requesting 10, making memory usage unpredictable.

3. Error Accumulation
   Iterative multiplication (v *= factor) accumulates errors across iterations, causing drift from true exponential spacing.

4. Floating point Prometheus Output
   Produces values like le="8.191999999999999e-05" instead of clean le="8.19e-05" due to floating-point representation issues.

5. No Edge Case Handling
   Can panic or produce incorrect results on invalid inputs (zero/negative buckets, start >= end).

Impact
From pprof analysis in performance and stability testing showing 39.91MB memory usage at /metrics endpoint:

66.40% (26.50MB) in prometheus.(*histogram).Write
33.83% (13.50MB) in prometheus.newHistogram
Root cause: Unpredictable bucket counts create more time series than expected, multiplying memory usage across all histograms.

(pprof) top
Showing nodes accounting for 64.31MB, 161.12% of 39.91MB total
Showing top 10 nodes out of 116
      flat  flat%   sum%        cum   cum%
   26.50MB 66.40% 66.40%    27.50MB 68.91%  github.com/prometheus/client_golang/prometheus.(*histogram).Write
   13.50MB 33.83% 100.23%    13.50MB 33.83%  github.com/prometheus/client_golang/prometheus.newHistogram
    6.66MB 16.68% 116.92%    25.16MB 63.04%  github.com/prometheus/client_golang/prometheus.(*metricMap).getOrCreateMetricWithLabels
    5.10MB 12.77% 129.68%    34.10MB 85.43%  github.com/prometheus/client_golang/prometheus.processMetric
    4.05MB 10.14% 139.82%     4.05MB 10.14%  google.golang.org/protobuf/proto.MarshalOptions.marshal
       3MB  7.52% 147.34%        3MB  7.52%  google.golang.org/protobuf/types/known/timestamppb.New (inline)
    1.50MB  3.76% 151.10%     1.50MB  3.76%  github.com/prometheus/client_golang/prometheus.populateMetric
    1.50MB  3.76% 154.86%     1.50MB  3.76%  encoding/json.(*decodeState).literalStore
    1.50MB  3.76% 158.62%     1.50MB  3.76%  github.com/prometheus/client_golang/prometheus.MakeLabelPairs
       1MB  2.51% 161.12%     4.50MB 11.28%  github.com/prometheus/client_golang/prometheus.v2.NewCounterVec.func1
(pprof) top -cum 
Showing nodes accounting for 0, 0% of 39.91MB total
Showing top 10 nodes out of 116
      flat  flat%   sum%        cum   cum%
         0     0%     0%    37.27MB 93.37%  github.com/hyperledger-labs/fabric-smart-client/platform/view/services/web/server/middleware.(*requestID).ServeHTTP
         0     0%     0%    37.27MB 93.37%  github.com/prometheus/client_golang/prometheus/promhttp.HandlerForTransactional.func1
         0     0%     0%    37.27MB 93.37%  github.com/prometheus/client_golang/prometheus/promhttp.InstrumentHandlerCounter.func1
         0     0%     0%    37.27MB 93.37%  github.com/prometheus/client_golang/prometheus/promhttp.InstrumentMetricHandler.InstrumentHandlerInFlight.func1
         0     0%     0%    37.27MB 93.37%  net/http.(*ServeMux).ServeHTTP
         0     0%     0%    37.27MB 93.37%  net/http.(*conn).serve
         0     0%     0%    37.27MB 93.37%  net/http.HandlerFunc.ServeHTTP
         0     0%     0%    37.27MB 93.37%  net/http.serverHandler.ServeHTTP
         0     0%     0%    34.10MB 85.43%  github.com/prometheus/client_golang/prometheus.(*Registry).Gather
         0     0%     0%    34.10MB 85.43%  github.com/prometheus/client_golang/prometheus.(*noTransactionGatherer).Gather

Solution
Implemented 5 fixes:

1. Safer calculation: Replaced math.Exp(math.Log(x)) with math.Pow(x, 1/n) - mathematically identical but clearer
2. Guaranteed exact bucket count: Changed loop to for i := 1; i < buckets; i++ ensuring exact count
3. Independent calculation: Each bucket calculated independently via math.Pow(factor, i) eliminating error accumulation
4. Clean Prometheus output: Added roundToSignificantDigits() producing values like le="0.00215443" instead of le="0.0021544299999999999"
5. Edge case handling: Returns sensible defaults for invalid inputs instead of panicking

Testing
Comprehensive test suite with individual tests for each fix:

✅ TestFix1_SaferCalculation_MaintainsExponentialSpacing
✅ TestFix2_GuaranteedExactBucketCount
✅ TestFix3_IndependentCalculation_NoErrorAccumulation
✅ TestFix4_RoundingProducesCleanValues
✅ TestFix5_EdgeCaseHandling

All tests passing with verification of:

• True exponential spacing (constant ratio between buckets)
• Exact bucket count (10 requested = 10 returned)
• Monotonically increasing values
• Clean Prometheus label format
• Robust edge case handling

Files Changed
platform/common/utils/metrics.go - Fixed implementation
platform/common/utils/metrics_test.go - Comprehensive test suite