fix(bitswap/httpnet): idempotent Stop()

[lidel]

This PR aims to avoid closing channel twice (inside of of stopCleaner) on shutdown.

Found in Kubo 0.35.0-rc2 (ipfs/kubo#10760), shutdown of ipfs daemon sometimes produces noisy panic (not a big deal, since this is shutdown, but unnecessary noise):

panic: close of closed channel

goroutine 16518 [running]:
github.com/ipfs/boxo/bitswap/network/httpnet.(*cooldownTracker).stopCleaner(...)
	github.com/ipfs/[email protected]/bitswap/network/httpnet/cooldown.go:50
github.com/ipfs/boxo/bitswap/network/httpnet.(*Network).Stop(0xc0014caa50)
	github.com/ipfs/[email protected]/bitswap/network/httpnet/httpnet.go:322 +0x2e
github.com/ipfs/boxo/bitswap/network.(*router).Stop(0xc0014fbb00)
	github.com/ipfs/[email protected]/bitswap/network/router.go:55 +0x35
github.com/ipfs/boxo/bitswap.(*Bitswap).Close(0xc0014fe100)
	github.com/ipfs/[email protected]/bitswap/bitswap.go:152 +0x22
github.com/ipfs/kubo/core/node.Online.ProvidingExchange.func4.1({0x2edbc40?, 0x39f9d20?})
	github.com/ipfs/kubo/core/node/bitswap.go:188 +0x1c
go.uber.org/fx/internal/lifecycle.(*Lifecycle).runStopHook(0xc000000460, {0x3a11130, 0x4f74900}, {0x0, 0xc0014941f8, {0x0, 0x0}, {0x0, 0x0}, {{0xc001580340, ...}, ...}})
	go.uber.org/[email protected]/internal/lifecycle/lifecycle.go:341 +0x1f2
go.uber.org/fx/internal/lifecycle.(*Lifecycle).Stop(0xc000000460, {0x3a11130, 0x4f74900})
	go.uber.org/[email protected]/internal/lifecycle/lifecycle.go:303 +0x52e
go.uber.org/fx.(*App).Stop.func2({0x3a11130?, 0x4f74900?})
	go.uber.org/[email protected]/app.go:725 +0x7a
go.uber.org/fx.withTimeout.func1()
	go.uber.org/[email protected]/app.go:803 +0x6b
created by go.uber.org/fx.withTimeout in goroutine 42
	go.uber.org/[email protected]/app.go:791 +0xc5

cc @hsanjuan for visibility, if there is a better way

[hsanjuan]

LGTM, but we close the network in two places then, which is probably unintended..

[lidel]

@hsanjuan could it be these two places? (once as bitswap, once as exchange)

• https://github.com/ipfs/kubo/blob/3290afcef8dba34296f26dfdc77554fd62dd3e8c/core/node/bitswap.go#L147
• https://github.com/ipfs/kubo/blob/3290afcef8dba34296f26dfdc77554fd62dd3e8c/core/node/bitswap.go#L163

I think its fine, my expectation for Close/Stop is to be idempotent anyway.