fix composite alert crash

[MisterSquishy]

fixes a latent bug in composite alerts, in which a drop-and-recreate of a subalert causes an OOB panic:

Stack trace from the terraform-provider-lightstep_v1.95.2 plugin:

panic: runtime error: index out of range [0] with length 0

goroutine 258 [running]:
github.com/lightstep/terraform-provider-lightstep/lightstep.buildCompositeAlert(0x12bd7b8?)
	github.com/lightstep/terraform-provider-lightstep/lightstep/resource_metric_condition.go:1109 +0x5e5
github.com/lightstep/terraform-provider-lightstep/lightstep.getUnifiedConditionAttributesFromResource(0x5b9a6a?, 0x1)
	github.com/lightstep/terraform-provider-lightstep/lightstep/resource_metric_condition.go:653 +0x1d9
github.com/lightstep/terraform-provider-lightstep/lightstep.(*resourceUnifiedConditionImp).resourceUnifiedConditionUpdate(0xc00035de88, {0xdb6048, 0xc0003e1b90}, 0xc0005bda00, {0xc8bcc0?, 0xc000114600?})
	github.com/lightstep/terraform-provider-lightstep/lightstep/resource_metric_condition.go:568 +0x94
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.(*Resource).update(0xc0004ccb60, {0xdb6080, 0xc00046a030}, 0xd?, {0xc8bcc0, 0xc000114600})
	github.com/hashicorp/terraform-plugin-sdk/[email protected]/helper/schema/resource.go:741 +0x12e
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.(*Resource).Apply(0xc0004ccb60, {0xdb6080, 0xc00046a030}, 0xc000e51450, 0xc0001e9480, {0xc8bcc0, 0xc000114600})
	github.com/hashicorp/terraform-plugin-sdk/[email protected]/helper/schema/resource.go:847 +0x845
github.com/hashicorp/terraform-plugin-sdk/v2/helper/schema.(*GRPCProviderServer).ApplyResourceChange(0xc000460d68, {0xdb6080?, 0xc000963e90?}, 0xc000d30140)
	github.com/hashicorp/terraform-plugin-sdk/[email protected]/helper/schema/grpc_provider.go:1021 +0xe8d
github.com/hashicorp/terraform-plugin-go/tfprotov5/tf5server.(*server).ApplyResourceChange(0xc0003c2320, {0xdb6080?, 0xc000962690?}, 0xc00037e070)
	github.com/hashicorp/[email protected]/tfprotov5/tf5server/server.go:818 +0x574
github.com/hashicorp/terraform-plugin-go/tfprotov5/internal/tfplugin5._Provider_ApplyResourceChange_Handler({0xc6b2a0?, 0xc0003c2320}, {0xdb6080, 0xc000962690}, 0xc00037e000, 0x0)
	github.com/hashicorp/[email protected]/tfprotov5/internal/tfplugin5/tfplugin5_grpc.pb.go:385 +0x170
google.golang.org/grpc.(*Server).processUnaryRPC(0xc0002ca1e0, {0xdb9a38, 0xc00012dd40}, 0xc0007a8000, 0xc000493290, 0x1280120, 0x0)
	google.golang.org/[email protected]/server.go:1336 +0xd33
google.golang.org/grpc.(*Server).handleStream(0xc0002ca1e0, {0xdb9a38, 0xc00012dd40}, 0xc0007a8000, 0x0)
	google.golang.org/[email protected]/server.go:1704 +0xa36
google.golang.org/grpc.(*Server).serveStreams.func1.2()
	google.golang.org/[email protected]/server.go:965 +0x98
created by google.golang.org/grpc.(*Server).serveStreams.func1
	google.golang.org/[email protected]/server.go:963 +0x28a

Error: The terraform-provider-lightstep_v1.95.2 plugin crashed!

this bug seems to have been latent since composite alert support was introduced in #144; we appear to have noticed it but it went unfixed.

this PR is not a perfect fix; it is still unclear to me why a change to hidden_queries triggers a drop-and-recreate of an entire alert object (but a change to, say, query_string does not), or why a drop-and-recreate would show up as an empty alert object in the list. but this seems like the quickest way out of this mess