Show C *callables* instead of "<native>" in Tachyon

[maurycy]

Shortly:

• no full C stack, only the frame
• numpy.dot (BLAS) etc. invisible,
• only CALL, CALL_KW and CALL_FUNCTION_EX
• no BINARY_OP etc.

This is the result:

maurycy@eiger cpython (tachyon-native-callable 32756e7) % ./python -m profiling.sampling run --native -d 3 /tmp/nat.py
Captured 3,001 samples in 3.00 seconds
Sample rate: 1,000.33 samples/sec
Error rate: 0.07
Profile Stats:
       nsamples   sample%   tottime (s)    cumul%   cumtime (s)  filename:lineno(function)
      2977/2977      99.3         2.977      99.3         2.977  ~:0(time.sleep)
          20/20       0.7         0.020       0.7         0.020  nat.py:7(<module>)
         1/2978       0.0         0.001      99.3         2.978  nat.py:3(cb)
         1/2979       0.0         0.001      99.3         2.979  ~:0(sorted)
         0/2979       0.0         0.000      99.3         2.979  nat.py:6(<module>)
         0/2999       0.0         0.000     100.0         2.999  ~:0(exec)
         0/2999       0.0         0.000     100.0         2.999  <frozen runpy>:87(_run_code)
         0/2999       0.0         0.000     100.0         2.999  <frozen runpy>:201(_run_module_as_main)
         0/2999       0.0         0.000     100.0         2.999  ~:0(<native>)

Legend:
  nsamples: Direct/Cumulative samples (direct executing / on call stack)
  sample%: Percentage of total samples this function was directly executing
  tottime: Estimated total time spent directly in this function
  cumul%: Percentage of total samples when this function was on the call stack
  cumtime: Estimated cumulative time (including time in called functions)
  filename:lineno(function): Function location and name
  
Summary of Interesting Functions:
  
Functions with Highest Direct/Cumulative Ratio (Hot Spots):
  1.000 direct/cumulative ratio, 99.3% direct samples: ~:(time.sleep)
  0.007 direct/cumulative ratio, 0.7% direct samples: nat.py:(<module>)
  0.000 direct/cumulative ratio, 0.0% direct samples: nat.py:(cb)
  
Functions with Highest Call Frequency (Indirect Calls):
  2999 indirect calls, 100.0% total stack presence: ~:(exec)
  2999 indirect calls, 100.0% total stack presence: <frozen runpy>:(_run_code)
  2999 indirect calls, 100.0% total stack presence: <frozen runpy>:(_run_module_as_main)

Functions with Highest Call Magnification (Cumulative/Direct):
  2979.0x call magnification, 2978 indirect calls from 1 direct: ~:(sorted)
  2978.0x call magnification, 2977 indirect calls from 1 direct: nat.py:(cb)
  149.9x call magnification, 2979 indirect calls from 20 direct: nat.py:(<module>)
maurycy@eiger cpython (tachyon-native-callable 32756e7) % 

for:

import time
def cb(x):
    time.sleep(0.005)
    return x
while True:
    sorted([2, 1], key=cb)
    "x" * 1000000

[maurycy]

We don't have debug offsets. The same comment around all offsetof occurences.

[maurycy]

@pablogsal Does this make directional sense to you? Please do not review this yet. This is mostly a proof concept. The prime problems are a) we don't have debug offsets here so it relies on elaborate offsetof - similarly in other places, b) this is only for callables.