fileio: start with 4k buffer for procfs

There's a very gradual increase of anonymous memory in systemd-journald that
blames to 2ac67221bb6270f0fbe7cbd0076653832cd49de2.

systemd-journald makes many calls to read /proc/PID/cmdline and
/proc/PID/status, both of which tend to be well under 4K. However the
combination of allocating 4M read buffers, then using `realloc()` to
shrink the buffer in `read_virtual_file()` appears to be creating
fragmentation in the heap (when combined with the other allocations
systemd-journald is doing).

To help mitigate this, try reading /proc with a 4K buffer as
`read_virtual_file()` did before 2ac67221bb6270f0fbe7cbd0076653832cd49de2.
If it isn't big enough then try again with the larger buffers.

src/basic/fileio.c

@@ -433,6 +433,11 @@ int read_virtual_file(const char *filename, size_t max_size, char **ret_contents
                         }
 
                         n_retries--;
+                } else if (n_retries > 1) {
+                        /* Files in /proc are generally smaller than the page size so let's start with a page size
+                         * buffer from malloc and only use the max buffer on the final try. */
+                        size = MIN3(page_size() - 1, READ_VIRTUAL_BYTES_MAX, max_size);
+                        n_retries = 1;
                 } else {
                         size = MIN(READ_VIRTUAL_BYTES_MAX, max_size);
                         n_retries = 0;