Fix incompatibility of BacklogQuota

[hezhangjian]

Motivation

there is a change between pulsar2.7 and master
command

bin/pulsar admin namespaces get-backlog-quotas $tenant/$namespace

pulsar2.7 returns

destination_storage    BacklogQuotaImpl(limit=102400, limitTime=-1, policy=consumer_backlog_eviction)

but pulsar 2.9 returns

"destination_storage    BacklogQuotaImpl(limitSize=102400, limitTime=-1, policy=consumer_backlog_eviction)"

So we need a compat, let old REST/sdk user, can set and get limit

Verifying this change

• add org.apache.pulsar.broker.resources.NamespaceBacklogQuotaCompatibleTest for test

Documentation

Check the box below and label this PR (if you have committer privilege).

Need to update docs?

• no-need-doc

It's a compatibility bug fix.

[codelipenghui]

@shoothzj I noticed the fix only fixes the compatibility issue when getting backlog quota, I think another compatibility issue is when using a new version client >= 2.8 but the broker which < 2.8, since the broker-side don't have limitSize and the client side will not pass the limit to the broker, so that the broker will use 0 as the value of value, it also should be a compatibility issue.

It's better to add a compatibility test to make sure we will not introduce breaking change for this part.

Thanks for the great work.

[hezhangjian]

@codelipenghui PTAL again, thanks

[hezhangjian]

@eolivelli @linlinnn PTAL, and should this cherry pick to 2.8 and 2.9 ? thanks

[eolivelli]

Are we breaking compatibility with 2.8 ?

2.7 is far away, and the current release is 2.8...
I wonder if it is better to restore compatibility with 2.7 and break against 2.8 or is it better to keep compatibility with 2.8 and add another point to the "breaking changes" ?

because IIUC we broke compatibility in 2.8 and now (we are at 2.9) it is too late

[eolivelli]

I have now understood better the final outcome of this patch.
I will cherry pick to branch-2.9 and prepare (again) 2.9.1