Flaky test: `TestBackoffRetryCanceledContext` in `exporters/otlp/otlplog/otlploggrpc`

[MrAlias]

make: *** [test/./exporters/otlp/otlplog/otlploggrpc] Error 1
--- FAIL: TestBackoffRetryCanceledContext (0.01s)
    retry_test.go:170: 
        	Error Trace:	/Users/runner/work/opentelemetry-go/opentelemetry-go/exporters/otlp/otlplog/otlploggrpc/internal/retry/retry_test.go:170
        	Error:      	Not equal: 
        	            	expected: 1
        	            	actual  : 2
        	Test:       	TestBackoffRetryCanceledContext
FAIL
FAIL	go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploggrpc/internal/retry	0.442s

[MrAlias]

example: https://github.com/open-telemetry/opentelemetry-go/actions/runs/19968993455/job/57268453009

[MrAlias]

The test is flaky because on slow systems it relies on the non-deterministic selection of the expected case statement:

1. fn is called
2. It returns an error
3. The code checks if the error is retryable, it always is
4. Time delay is checked
   • Max elsapsed time is 10 ms
   • Initial delay is 1ms
   • Delay is determined to be 1ms
   • The program proceeds to waiting
5. Wait is called
6. The wait select statement is evaluated
   • On slow systems both cases are true
     • Non-deterministically the timer channel case is selected
     • The retry function is re-run and and second iteration is recorded causing the failure
   • On fast systems only the context cancel is true
     • The retry stops here with only 1 execution

[MrAlias]

Proposed fix: check the context cancellation before any evaluation of the delay and performing a wait.