Incorrect operation time computation if variable baseline measurement

[petrs]

The computation of time of single operation is incorrectly computed if the 'baseline measurements ' or 'baseline stats' is not taking same time (+ noise) - for example, if additional garbage collection is executed during otherwise identical baseline measurements command, the results after average computation will be biased.

Possible solution:

• if outlayer measurement is detected, then increase the number of measurements repetitions (only for this operation), drop outlayers and compute average only from the rest

Example:
(https://github.com/crocs-muni/jcalgtest_results/algtest_results/javacard/Profiles/performance/fixed/Infineon_SECORA_ID_X_Batch_16072021_SALES____PERFORMANCE_SYMMETRIC_ASYMMETRIC_DATAFIXED_1627630268599__3b_88_80_01_00_00_00_11_77_81_c3_00_2d_(provided_by_Thoth).csv)

method name:; TYPE_RSA_PUBLIC LENGTH_RSA_1024 ALG_RSA_NOPAD Cipher_doFinal()
measurement config:;appletPrepareINS;31;appletMeasureINS;43;config;00 11 00 0c 00 01 00 04 04 00 00 07 00 80 ff ff 00 02 00 32 00 01
baseline measurements (ms):;10179.00;5296.00;5296.00;5297.00;5292.00;
baseline stats (ms):;avg:;6272.00;min:;5292.00;max:;10179.00;;;CHECK
operation raw measurements (ms):;-651.00;-657.00;-655.00;-655.00;-655.00;
operation stats (ms/op):;avg op:;-13.09;min op:;-13.14;max op:;-13.02;
operation info:;data length;128;total iterations;250;total invocations;250;