[Issue][hipblaslt] nan values on m=1,k=1 kernels.

[carsonbrownlee]

issue:
hipblaslt-bench is reporting norm_error NAN for GEMM size m=1,k=1. Tested on cases of S, BSS, S_MX, and S_BSS and multiple transposes on mi350. Tested on latest develop 44ad2d4. Possibly error with norm_error calculation?

to reproduce:
hipblaslt-bench --batch_count 1 -m 1 -n 2 -k 1 --stride_a 0 --stride_b 0 --stride_c 0 --stride_d 0 --alpha 1 --beta 0 --transA N --transB N --scaleA 0 --scaleB 0 --bi
as_source d --flush --rotating 512 --initialization trig_float --print_kernel_info --any_stride -i 10000 -j 24000 --a_type bf16_r --b_type bf16_r --c_type b
f16_r --d_type bf16_r --compute_type f32_r --verify

N,N,0,1,1,2,1,1,1,0,0,1,0,1,0,1,0,bf16_r,bf16_r,bf16_r,bf16_r,f32_r,0,0,0,0,0,0,0,none,0,bf16_r,bf16_r,512,1,0, <redacted> ,nan,1e-06,1e-06
--Solution index: 4975
--Solution name:  Cijk_Ailk_Bljk_BBS_BH_Bias_HA_S_SAV_UserArgs_MT1x2x512_SN_LDSB0_AFC1_AFEM1_AFEM1_ASEM1_CLR0_CADS0_DTLA0_DTLB0_DTVA0_DTVB0_EPS0_FDSI0_GRPM1_GRVWA1_GRVWB1_GSU1_GSUAMB_GSUC0_GSUWGMRR0_GLS0_ISA950_IU1_K1_LBSPPA0_LBSPPB0_LBSPPM0_LPA2_LPB8_LPM0_LRVW2_LWPMn1_MIAV0_MO40_NTn1_NTA0_NTB0_NTC0_NTD0_NTM0_NEPBS0_NLCA1_NLCB8_ONLL1_PGR0_PLR0_PKA1_SIA1_SS0_SU32_SUM0_SUS1024_SPO0_SRVW0_SSO0_SVW1_SK0_SKXCCM0_TT1_2_TLDS2_ULSGRO0_USL1_UIOFGRO0_USFGROn1_VSn1_VWA1_VWB1_WSGRA0_WSGRB0_WS64_WG1_1_64_WGM1_WGMXCC1_WGMXCCGn1
--kernel name:    Cijk_Ailk_Bljk_BBS_BH_Bias_HA_S_SAV_UserArgs_MT1x2x512_SN_LDSB0_AFC1_AFEM1_AFEM1_ASEM1_CLR0_CADS0_DTLA0_DTLB0_DTVA0_DTVB0_EPS0_FDSI0_GRPM1_GRVWA1_GRVWB1_GSUAMB_GLS0_ISA950_IU1_K1_LBSPPA0_LBSPPB0_LBSPPM0_LPA2_LPB8_LPM0_LRVW2_LWPMn1_MIAV0_MO40_NTn1_NTA0_NTB0_NTC0_NTD0_NTM0_NEPBS0_NLCA1_NLCB8_ONLL1_PGR0_PLR0_PKA1_SIA1_SS0_SPO0_SRVW0_SSO0_SVW1_SK0_SKXCCM0_TT1_2_TLDS2_ULSGRO0_USL1_UIOFGRO0_USFGROn1_VSn1_VWA1_VWB1_WSGRA0_WSGRB0_WS64_WG1_1_64