zfsslower (bcc version) does not handle zpl_iter_{read,write} correctly

In tools/zfsslower.py:

```c
// zpl_read(), zpl_write():
int trace_rw_entry(struct pt_regs *ctx, struct file *filp, char __user *buf,
size_t len, loff_t *ppos)
```

```python
if BPF.get_kprobe_functions(b'zpl_iter.*'):
b.attach_kprobe(event="zpl_iter_read", fn_name="trace_rw_entry")
b.attach_kprobe(event="zpl_iter_write", fn_name="trace_rw_entry")
```

However, in zfs code:

https://github.com/openzfs/zfs/blob/0a4b5976542e794c3231b0c5c3d504903e392250/module/os/linux/zfs/zpl_file.c#L215

```c
static ssize_t
zpl_iter_read(struct kiocb *kiocb, struct iov_iter *to)
```

The function signature does not match, making zfsslower printing misleading output. The libbpf version fsslower does not have this issue.

A possible fix:

```c
// zpl_iter_read(), zpl_iter_write():
int trace_zpl_iter_rw_entry(struct pt_regs *ctx,
struct kiocb *kiocb,
struct iov_iter *iter)
{
u64 id = bpf_get_current_pid_tgid();
u32 tgid = id >> 32;

if (FILTER_PID)
return 0;

if (!kiocb)
return 0;

struct val_t val = {};
val.ts = bpf_ktime_get_ns();

bpf_probe_read_kernel(&val.fp, sizeof(val.fp), &kiocb->ki_filp);
bpf_probe_read_kernel(&val.offset, sizeof(val.offset), &kiocb->ki_pos);

if (val.fp)
entryinfo.update(&id, &val);

return 0;
}
```

And use this as kprobe for iter instead.

Start in tools/zfsslower.py by comparing the existing trace_rw_entry handler with the zpl_iter_read and zpl_iter_write signatures shown in the issue. Trace how the iter probes are attached and how entryinfo is used. Done means the iter probes use a matching handler and zfsslower no longer prints misleading output for those functions.

Written by the indexing model from the issue text.