fix: assign new_nodes of service discovery to upstream every time

[nic-6443]

Description

I introduced a performance optimization in #12258, and after testing, we found that the kubernetes performance optimization would invalid after some time.
The reason is that after watch progress is done in kubernetes service discovery for some time, it will re-list all data. At this point, addresses of all tables that store service discovery are changed, but the content of nodes does not change. This causes up_conf.nodes to never reference the new_nodes table at

apisix/apisix/upstream.lua

Lines 305 to 317 in 9e5163c

	local same = upstream_util.compare_upstream_node(up_conf, new_nodes)
	if not same then
	if not up_conf._nodes_ver then
	up_conf._nodes_ver = 0
	end
	up_conf._nodes_ver = up_conf._nodes_ver + 1
	local pass, err = core.schema.check(core.schema.discovery_nodes, new_nodes)
	if not pass then
	return HTTP_CODE_UPSTREAM_UNAVAILABLE, "invalid nodes format: " .. err
	end
	up_conf.nodes = new_nodes

because their addresses are different and it goes to the branch where equality is determined by value in every downstream requests.
This branch has significant performance overhead:

apisix/apisix/utils/upstream.lua

Lines 50 to 61 in 9e5163c

	core.table.sort(old_t, sort_by_key_host)
	core.table.sort(new_t, sort_by_key_host)
	for i = 1, #new_t do
	local new_node = new_t[i]
	local old_node = old_t[i]
	for _, name in ipairs({"host", "port", "weight", "priority", "metadata"}) do
	if new_node[name] ~= old_node[name] then
	return false
	end
	end
	end

Checklist

• I have explained the need for this PR and the problem it solves
• I have explained the changes or the new features added to this PR
• I have added tests corresponding to this change
• I have updated the documentation to reflect this change
• I have verified that this change is backward compatible (If not, please discuss on the APISIX mailing list first)

[ronething]

lgtm

[nic-chen]

Call delay_encode once for each log record, otherwise the content of the log records will be garbled.

Suggested change

	core.log.debug("compare upstream nodes by value, ",
	"old: ", tostring(old_t) , " ", core.json.delay_encode(old_t, true),
	", new: ", tostring(new_t) , " ", core.json.delay_encode(new_t, true))
	core.log.debug("compare upstream nodes by value, ",
	"old: ", tostring(old_t) , " ", core.json.delay_encode(old_t, true))
	core.log.debug(" new: ", tostring(new_t) , " ", core.json.delay_encode(new_t, true))

[nic-6443]

fixed