pgrac-start: false "node does not match" warning when pgrac.conf is shared across nodes

[sqlrush]

Summary

pgrac-start preflight warns when the first [node.N] section in
pgrac.conf doesn't match this node's cluster.node_id. But the recommended
setup uses one identical pgrac.conf shared by every node, so on any node
whose id isn't the first-listed one the warning fires even though the config is
correct and startup succeeds.

Current behavior

With a shared pgrac.conf listing [node.0] then [node.1], starting node 1
prints:

pgrac-start: WARNING: /path/pgrac.conf first [node.0] does not match postgresql.conf cluster.node_id=1
pgrac-start: WARNING: postmaster will FATAL on startup unless cluster.node_id appears as a [node.N] in pgrac.conf.  Run pgrac-init --force to fix.

…then the server starts fine, because node 1 is declared (as [node.1]). The
warning is a false positive and the suggested pgrac-init --force is misleading.

Desired behavior

Warn only on the real failure: cluster.node_id does not appear as
any [node.N] section. When the id is present (in any position), don't warn.

Why this is a good first issue

• It's in a bash script (src/bin/pgrac/pgrac-start), not the C backend.
• Current logic reads only the first [node.N]; the fix is to scan for all
  declared node ids and test membership.
• There's already a TAP test to extend: src/bin/pgrac/t/002_start.pl.

Definition of done

• No warning when cluster.node_id is declared as some [node.N], regardless
  of order.
• Still warns clearly when the id is genuinely absent.
• A regression test in t/002_start.pl: a shared multi-node pgrac.conf
  started as a non-first node asserts no warning.

Pointers

• src/bin/pgrac/pgrac-start — preflight that reads the first [node.N] and
  emits the WARNING: ... first [node. lines.
• src/bin/pgrac/t/002_start.pl — existing start tests.