Concurrent DELETE download jobs crashes the target

Is there an existing issue for this?

• I have searched the existing issues

Describe the bug

To reproduce, generate some download job history, and run concurrent requests to delete some of them (through the DELETE /v1/download/remove endpoint)

Expected Behavior

• Ideally requests are handled concurrently
• Or less ideally some requests are failed but retryable, without crashing the server application

Current Behavior

Target pod crashes with

fatal error: concurrent map writes

goroutine 2205552 [running]:
internal/runtime/maps.fatal({0x34d84f3?, 0x2f73540?})
	runtime/panic.go:1181 +0x18
internal/runtime/maps.(*Map).Delete(0x19d6f4088390, 0x2f73540, 0x19d6f3d19470)
	internal/runtime/maps/map.go:682 +0x125
github.com/NVIDIA/aistore/ext/dload.(*infoStore).delJob(...)
	github.com/NVIDIA/aistore/ext/dload/infostore.go:126
github.com/NVIDIA/aistore/ext/dload.(*dispatcher).handleRemove(0x36f4da0?, 0x19d6f3d19688)
	github.com/NVIDIA/aistore/ext/dload/dispatcher.go:383 +0xfc
github.com/NVIDIA/aistore/ext/dload.(*dispatcher).adminReq(0x19d6f4e05400, 0x19d6f3d19688)
	github.com/NVIDIA/aistore/ext/dload/dispatcher.go:365 +0x148
github.com/NVIDIA/aistore/ext/dload.(*Xact).RemoveJob(0x19d6f1fa2480, {0x19d6f6c008f0?, 0x34b03eb?})
	github.com/NVIDIA/aistore/ext/dload/xact.go:244 +0x85
github.com/NVIDIA/aistore/ais.(*target).downloadHandler(0x19d6f1fa2180, {0x36d5008, 0x19d6f3e064b0}, 0x19d6f3d1c640)
	github.com/NVIDIA/aistore/ais/tgtdl.go:178 +0xb1b

Steps To Reproduce

As mentioned above, happy to provide a more step by step script if needed.

Possible Solution

Seems like a mutex fix needed.

Additional Information/Context

No response

AIStore build/version

v5.0

Environment details (OS name and version, etc.)

Running in kubernetes with the recommended operator deployment charts.

Start with ext/dload/infostore.go at infoStore.delJob, then trace the DELETE /v1/download/remove path through dispatcher.go, xact.go, and ais/tgtdl.go. Reproduce concurrent deletion of download jobs and verify that requests no longer crash the target, while any rejected requests remain retryable.