Fix HostDB sync issue by preserving expiry_time on load

[yknoya]

Problem

Although PR #12395 addresses an issue related to loading host.db, a new issue has been identified concerning the synchronization to host.db after the fix.
In short, the cache loaded from host.db is not being synced back to host.db.

Since synchronization to host.db is handled by HostDBSync, I investigated the behavior and found that the write_partition method in HostDBSync was skipping the write operation.
Specifically, the following if statement was triggered, and according to a gdb inspection, entry->meta.expiry_time was set to -1:

trafficserver/iocore/hostdb/P_RefCountCacheSerializer.h

Lines 167 to 170 in 9f6abab

	// check if the item has expired, if so don't persist it to disk
	if (entry->meta.expiry_time < curr_time) {
	continue;
	}

(gdb) p entry->meta.expiry_time
$1 = -1
(gdb) p curr_time
$2 = 1754462975

Cause

The cause of entry->meta.expiry_time being -1 lies in the load process of host.db, specifically in the LoadRefCountCacheFromPath function:

trafficserver/iocore/hostdb/P_RefCountCache.h

Lines 612 to 615 in 9f6abab

	CacheEntryType *newItem = load_func((char *)&buf, tmpValue.size);
	if (newItem != nullptr) {
	cache.put(tmpValue.key, newItem, tmpValue.size - sizeof(CacheEntryType));
	}

The cache is registered via a call to cache.put (RefCountCache<HostDBInfo>::put), but since the 4th argument (expiry_time) is omitted, the default value -1 is used:

trafficserver/iocore/hostdb/P_RefCountCache.h

Line 405 in 9f6abab

void put(uint64_t key, C *item, int size = 0, ink_time_t expiry_time = -1);

Fix

The correct value to pass for the 4th argument of RefCountCache<HostDBInfo>::put is stored in tmpValue.expiry_time.
This patch updates the call to explicitly pass tmpValue.expiry_time as the expiry_time argument.

[ezelkow1]

[approve ci autest 0]

[ezelkow1]

[approve ci]