[SYCL][L0] Add XPTI-based Level Zero tracing

[alexbatashev]

Add support for tracing Level Zero API calls to Level Zero plugin. Also modify sycl-trace tool to be able to display both PI and L0 calls simultaneously.

TBD tests link

Example PI-only trace:

[PI] piMemBufferCreate(
    context: 0x1511b60
    flags: 1
    size: 16
    host_ptr: 0
    ret_mem: 0x7ffceb731460
    properties: 0
) ---> PI_SUCCESS
[PI] piextMemGetNativeHandle(
    mem: 0x16c1130
    nativeHandle: 0x7ffceb731268
) ---> PI_SUCCESS
[PI] piDeviceGetInfo(
    device: 0xda40f0
    param_name: 65576
    param_value_size: 4
    param_value: 0x7ffceb730e24
    param_value_size_ret: 0
) ---> PI_SUCCESS
[PI] piextDeviceSelectBinary(
    device: 0xda40f0
    binaries: 0x16c02c0
    num_binaries: 1
    selected_binary_ind: 0x7ffceb730d14
) ---> PI_INVALID_BINARY

Example of verbose output:

[PI:TID 140400402187264:<unknown>:0]
piEnqueueMemBufferMap(
    command_queue: 0x13aae70
    buffer: 0x160da50
    blocking_map: 0
    map_flags: 1
    offset: 0
    size: 16
    num_events_in_wait_list: 1
    event_wait_list: 0x20b08c0
    event: 0x1eb7758
    ret_map: 0x7ffeba138a70
) {
*  [L0:TID 140400402187264:<unknown>:0]
*  zeEventQueryStatus(
*     phEvent: 0x7ffeba1385f8
*  ) ---> ZE_RESULT_NOT_READY (synchronization primitive not signaled)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeEventCreate(
*     phEventPool: 0x7ffeba1385a8
*     pdesc: 0x7ffeba1385a0
*     pphEvent: 0x7ffeba138598
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeEventHostSynchronize(
*     phEvent: 0x7ffeba1385e8
*     ptimeout: 0x7ffeba1385e0
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeFenceQueryStatus(
*     phFence: 0x7ffeba138578
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeFenceReset(
*     phFence: 0x7ffeba138528
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeCommandListReset(
*     phCommandList: 0x7ffeba138528
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
*  [L0:TID 140400402187264:<unknown>:0]
*  zeEventHostSignal(
*     phEvent: 0x7ffeba138648
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
} ---> PI_SUCCESS

[PI:TID 140400402187264:<unknown>:0]
piEventsWait(
    num_events: 1
    event_list: 0x1d7ef90
) {
*  [L0:TID 140400402187264:<unknown>:0]
*  zeEventHostSynchronize(
*     phEvent: 0x7ffeba138928
*     ptimeout: 0x7ffeba138920
*  ) ---> ZE_RESULT_SUCCESS (success)
*  
} ---> PI_SUCCESS

[alexbatashev]

Example SYCL performance trace

simple.txt (replace txt with json)

[smaslov-intel]

Continuing in #6023