Retry on rate-limiting could be smarter

[tonistiigi]

Server sends some extra headers with 429 as visible in docker/buildx#841 (comment)

2021-11-25T11:07:28.4672333Z 2021/11/25 11:07:26 < response 429 Too Many Requests map[Activityid:[3db07705-7b00-43ef-b1bf-d9d6ce0d8a44] Content-Length:[315] Content-Type:[application/json] Date:[Thu, 25 Nov 2021 11:07:26 GMT] Retry-After:[60] Strict-Transport-Security:[max-age=2592000] X-Cache:[CONFIG_NOCACHE] X-Frame-Options:[SAMEORIGIN] X-Msedge-Ref:[Ref A: A66EBD35FE94468991D34EC956C93CBB Ref B: ASHEDGE1519 Ref C: 2021-11-25T11:07:26Z] X-Ratelimit-Limit:[493] X-Ratelimit-Remaining:[0] X-Ratelimit-Reset:[1637838743] X-Ratelimit-Resource:[ArtifactCache/Short] X-Tfs-Processid:[2ef82cec-56f3-4c18-b269-d6d60d120383] X-Tfs-Serviceerror:[Request+was+blocked+due+to+exceeding+usage+of+resource+%27Count%27+in+namespace+%27%27.] X-Tfs-Session:[3db07705-7b00-43ef-b1bf-d9d6ce0d8a44] X-Vss-E2eid:[3db07705-7b00-43ef-b1bf-d9d6ce0d8a44] X-Vss-Senderdeploymentid:[5f814983-29e7-4f0f-d4b8-8016a66b281b]]

We could use Retry-After, X-Ratelimit-Reset for smarter timeout between retries instead of the current one that tries to detect it on client side.