Fix bug / leak in JS function processors (elastic#12600) (elastic#12604)

In an effort to minimize allocations a slice used to hold function arguments is reused rather than allocation new each time a Javascript function processor is used. However there was a typo in re-slicing code so instead of reseting the length to 0 it was a no-op. Hence all future invocations appended to the slice causing a leak. This affects code that uses native Javascript functions in a processor chain (e.g. `new process.Chain().Add(function(event) { event.Put("x", "y"); }).Build()`).

(cherry picked from commit f69330a)

CHANGELOG.asciidoc

@@ -66,6 +66,7 @@ https://github.com/elastic/beats/compare/v7.1.1...v7.2.0[View commits]
 - Add host.os.codename to fields.yml. {pull}12261[12261]
 - Fix `@timestamp` being duplicated in events if `@timestamp` is set in a
   processor (or by any code utilizing `PutValue()` on a `beat.Event`).
+- Fix leak in script processor when using Javascript functions in a processor chain. {pull}12600[12600]
 
 *Auditbeat*
 

libbeat/processors/script/javascript/module/processor/chain.go

@@ -83,7 +83,7 @@ func (b *chainBuilder) Add(call goja.FunctionCall) goja.Value {
 	case *beatProcessor:
 		b.procs = append(b.procs, v.p)
 	case func(goja.FunctionCall) goja.Value:
-		b.procs = append(b.procs, &jsProcessor{fn: v})
+		b.procs = append(b.procs, newJSProcessor(v))
 	default:
 		panic(b.runtime.NewGoError(errors.Errorf("arg0 must be a processor object, but got %T", a0.Export())))
 	}
@@ -119,9 +119,12 @@ type jsProcessor struct {
 	call goja.FunctionCall
 }
 
+func newJSProcessor(fn jsFunction) *jsProcessor {
+	return &jsProcessor{fn: fn, call: goja.FunctionCall{Arguments: make([]goja.Value, 1)}}
+}
+
 func (p *jsProcessor) run(event javascript.Event) error {
-	p.call.Arguments = p.call.Arguments[0:]
-	p.call.Arguments = append(p.call.Arguments, event.JSObject())
+	p.call.Arguments[0] = event.JSObject()
 	p.fn(p.call)
 	return nil
 }