Fix process hog metrics

Some of them were mathematically incorrect, for example:

proc.hog.cpu = 100 * (rate(proc.psinfo.utime) + rate(proc.psinfo.stime)) / (kernel.all.uptime - proc.psinfo.start_time)

This decays over time since the denominator grows while
the numerator has a maximum limit.

Fix the cpu/disk/net formulas to match the descriptions.

Make help texts slightly more clearer and consistent
while at it.

Resolves: #2391

Author: myllynen

src/derived/proc.conf

@@ -4,33 +4,33 @@
 proc.psinfo.age = kernel.all.uptime - proc.psinfo.start_time
 proc.psinfo.age(oneline) = time in seconds since process was started
 proc.psinfo.age(helptext) = '\
-time in seconds since process was started, calculated from
-proc.psinfo.start_time subtracted from kernel.all.uptime.'
+time in seconds since process was started, calculated
+as kernel.all.uptime minus proc.psinfo.start_time.'
 
-proc.io.total_bytes = instant(proc.io.read_bytes) + instant(proc.io.write_bytes) - instant(proc.io.cancelled_write_bytes)
+proc.io.total_bytes = instant(proc.io.read_bytes + proc.io.write_bytes - proc.io.cancelled_write_bytes)
 proc.io.total_bytes(oneline) = total bytes read and written by process
 proc.io.total_bytes(helptext) = '\
-total bytes read and written by the process, less cancelled written bytes.'
+total bytes read and written by process, minus cancelled written bytes.'
 
 # metrics to identify resource hogs
-proc.hog.cpu = 100 * (rate(proc.psinfo.utime) + rate(proc.psinfo.stime)) / (kernel.all.uptime - proc.psinfo.start_time)
-proc.hog.cpu(oneline) = average percentage CPU utilization of each process
+proc.hog.cpu = 100 * instant(proc.psinfo.utime + proc.psinfo.stime) / (kernel.all.uptime - proc.psinfo.start_time)
+proc.hog.cpu(oneline) = average CPU utilization of process
 proc.hog.cpu(helptext) = '\
-average CPU utilization of each process expressed as a percentage
-of time since the process started.'
+average CPU utilization of process expressed as
+a percentage of time since process start.'
 
 proc.hog.mem = proc.memory.vmrss + proc.memory.vmswap
-proc.hog.mem(oneline) = sum of resident and swapped memory used by the process
+proc.hog.mem(oneline) = sum of resident and swapped memory used by process
 proc.hog.mem(helptext) = '\
-amount of resident and swapped memory used by each process.'
+sum of resident and swapped memory used by process.'
 
-proc.hog.disk = (rate(proc.io.read_bytes) + rate(proc.io.write_bytes) - rate(proc.io.cancelled_write_bytes)) / (kernel.all.uptime - proc.psinfo.start_time)
-proc.hog.disk(oneline) = average I/O rate of each process since it was started
+proc.hog.disk = instant(proc.io.read_bytes + proc.io.write_bytes - proc.io.cancelled_write_bytes) / (kernel.all.uptime - proc.psinfo.start_time)
+proc.hog.disk(oneline) = average I/O rate of process
 proc.hog.disk(helptext) = '\
-average I/O rate (reads and writes less cancelled writes) of each process
-since it was started.'
+average I/O rate of process (reads and writes minus cancelled writes)
+since process start.'
 
-proc.hog.net = rate(bpf.proc.net.tcp.recv.bytes) + rate(bpf.proc.net.tcp.send.bytes) + rate(bpf.proc.net.udp.recv.bytes) + rate(bpf.proc.net.udp.send.bytes)
-proc.hog.net(oneline) = sum of network usage by the process
+proc.hog.net = rescale(instant(bpf.proc.net.tcp.recv.bytes + bpf.proc.net.tcp.send.bytes + bpf.proc.net.udp.recv.bytes + bpf.proc.net.udp.send.bytes), "Kbyte")
+proc.hog.net(oneline) = sum of network usage by process
 proc.hog.net(helptext) = '\
-sum of network usage (TCP/UDP recv/send) by each process.'
+sum of network usage (TCP/UDP recv/send) by process.'